{ "input": { "test_number": 11 }, "steps": { "input": { "test_number": 11 }, "build-batches": { "rwBatches": [ { "mod": 1, "type": "vocabulary_in_context", "count": 5, "startQ": 1, "testNumber": 11 }, { "mod": 1, "type": "text_comprehension", "count": 8, "startQ": 6, "testNumber": 11 }, { "mod": 1, "type": "data_analysis", "count": 4, "startQ": 14, "testNumber": 11 }, { "mod": 1, "type": "claim_support", "count": 4, "startQ": 18, "testNumber": 11 }, { "mod": 1, "type": "standard_english_conventions", "count": 5, "startQ": 22, "testNumber": 11 }, { "mod": 1, "type": "transitions", "count": 4, "startQ": 27, "testNumber": 11 }, { "mod": 1, "type": "rhetorical_synthesis", "count": 3, "startQ": 31, "testNumber": 11 }, { "mod": 2, "type": "vocabulary_in_context", "count": 5, "startQ": 1, "testNumber": 11 }, { "mod": 2, "type": "text_comprehension", "count": 8, "startQ": 6, "testNumber": 11 }, { "mod": 2, "type": "data_analysis", "count": 4, "startQ": 14, "testNumber": 11 }, { "mod": 2, "type": "claim_support", "count": 4, "startQ": 18, "testNumber": 11 }, { "mod": 2, "type": "standard_english_conventions", "count": 5, "startQ": 22, "testNumber": 11 }, { "mod": 2, "type": "transitions", "count": 4, "startQ": 27, "testNumber": 11 }, { "mod": 2, "type": "rhetorical_synthesis", "count": 3, "startQ": 31, "testNumber": 11 } ], "mathBatches": [ { "mod": 1, "type": "algebra", "count": 7, "spr": 2, "startQ": 1, "testNumber": 11 }, { "mod": 1, "type": "functions_analysis", "count": 5, "spr": 1, "startQ": 8, "testNumber": 11 }, { "mod": 1, "type": "geometry_trigonometry", "count": 5, "spr": 1, "startQ": 13, "testNumber": 11 }, { "mod": 1, "type": "statistics_probability", "count": 5, "spr": 1, "startQ": 18, "testNumber": 11 }, { "mod": 1, "type": "advanced_math", "count": 5, "spr": 2, "startQ": 23, "testNumber": 11 }, { "mod": 2, "type": "algebra", "count": 7, "spr": 2, "startQ": 1, "testNumber": 11 }, { "mod": 2, "type": "functions_analysis", "count": 5, "spr": 1, "startQ": 8, "testNumber": 11 }, { "mod": 2, "type": "geometry_trigonometry", "count": 5, "spr": 1, "startQ": 13, "testNumber": 11 }, { "mod": 2, "type": "statistics_probability", "count": 5, "spr": 1, "startQ": 18, "testNumber": 11 }, { "mod": 2, "type": "advanced_math", "count": 5, "spr": 2, "startQ": 23, "testNumber": 11 } ], "testNumber": 11 }, "gen-rw": { "count": 14, "costUsd": 2.06721, "failed": 0, "items": [ { "costUsd": 0.084615, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Reading & Writing questions.\nType: vocabulary_in_context | Test: 11 | Unit: 1 | Start Q#: 1\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m1-qN\" where N is the sequential question number starting at 1.\nFor example, the first question in this batch has id \"sat11-rw-m1-q1\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m1-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m1-q1\",\"number\":1,\"type\":\"vocabulary_in_context\",\"passage\":\"In the decades following the publication of Keynes's General Theory, economists were reluctant to ______ the orthodoxy of classical market equilibrium outright. Instead, they tended to graft Keynesian insights onto existing models, preserving much of the older framework even as its foundational assumptions were being quietly eroded.\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"repudiate\",\"B\":\"disregard\",\"C\":\"misconstrue\",\"D\":\"lament\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m1-q2\",\"number\":2,\"type\":\"vocabulary_in_context\",\"passage\":\"Although Giorgio Vasari praised Pontormo's late frescoes for their technical daring, later critics found the figures' elongated limbs and unnatural palette deeply ______. To these observers, the Mannerist departure from High Renaissance proportion signaled not innovation but a troubling loss of classical composure.\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"prosaic\",\"B\":\"derivative\",\"C\":\"disconcerting\",\"D\":\"ostentatious\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m1-q3\",\"number\":3,\"type\":\"vocabulary_in_context\",\"passage\":\"Behavioral economists have long observed that consumers do not weigh gains and losses symmetrically; a loss of \\\\$100 tends to ______ the pleasure of a \\\\$100 gain. This asymmetry, known as loss aversion, helps explain why investors often hold losing stocks far longer than rational models predict.\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"eclipse\",\"B\":\"replicate\",\"C\":\"supplement\",\"D\":\"anticipate\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m1-q4\",\"number\":4,\"type\":\"vocabulary_in_context\",\"passage\":\"Because the faint signal from the distant quasar was nearly ______ by the glow of intervening galaxies, astronomers at the Keck Observatory employed adaptive optics to isolate its spectrum. Only after filtering out the foreground light could they confirm the object's redshift with any confidence.\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"amplified\",\"B\":\"refracted\",\"C\":\"subsumed\",\"D\":\"occluded\"},\"answer_format\":\"mcq\",\"correct_answer\":\"D\"},{\"id\":\"sat11-rw-m1-q5\",\"number\":5,\"type\":\"vocabulary_in_context\",\"passage\":\"Hannah Arendt argued that totalitarian regimes do not merely suppress dissent; they seek to ______ the very categories through which citizens distinguish truth from falsehood. Once this shared epistemic ground collapses, she warned, resistance itself becomes almost unintelligible.\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"codify\",\"B\":\"dissolve\",\"C\":\"arbitrate\",\"D\":\"venerate\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}]" }, { "costUsd": 0.25419, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 8 SAT Reading & Writing questions.\nType: text_comprehension | Test: 11 | Unit: 1 | Start Q#: 6\n\nMANDATORY: Your JSON array MUST contain exactly 8 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m1-qN\" where N is the sequential question number starting at 6.\nFor example, the first question in this batch has id \"sat11-rw-m1-q6\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m1-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 8 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m1-q6\",\"number\":6,\"type\":\"text_comprehension\",\"passage\":\"The following is adapted from Willa Cather's 1918 novel My Ántonia. The narrator recalls his arrival at his grandparents' Nebraska farm as a child.\\n\\nI had the sense of coming home to myself, and of having found out what a little circle man's experience is. For the first time, perhaps, since that night, I was dissolved into something complete and great. When it came to me, it came as naturally as sleep. The wind sang a sort of clear tune through the long grass, and I lay down among the pumpkins and did not try to think of anything.\",\"prompt\":\"Which choice best describes the narrator's state of mind in the passage?\",\"choices\":{\"A\":\"He feels a profound sense of peace that quiets his impulse toward reflection.\",\"B\":\"He is disoriented by the unfamiliar landscape and seeks refuge among the pumpkins.\",\"C\":\"He is eager to explore the farm but exhausted from his long journey.\",\"D\":\"He is nostalgic for a former home he has recently left behind.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m1-q7\",\"number\":7,\"type\":\"text_comprehension\",\"passage\":\"In a 2019 study, ecologist Mariana Altrichter examined the role of white-lipped peccaries in the Gran Chaco forests of Paraguay. Altrichter found that the peccaries, which travel in herds of up to three hundred, create wallows—shallow muddy depressions—that persist long after the herds have moved on. These wallows retain water during dry seasons and become refuges for frogs, insects, and small mammals that would otherwise perish. When peccary populations collapse due to hunting, the wallows dry up within a single season, and local biodiversity declines sharply.\",\"prompt\":\"Which choice best states the main idea of the text?\",\"choices\":{\"A\":\"Peccary herds in the Gran Chaco have grown in size as hunting pressures have decreased in recent decades.\",\"B\":\"Altrichter's 2019 study was the first to document the behavior of white-lipped peccaries in Paraguay.\",\"C\":\"The disappearance of peccaries threatens the frogs and insects that share their habitat in the Gran Chaco.\",\"D\":\"By creating wallows that sustain other species, peccaries function as ecosystem engineers whose loss reverberates through the Chaco food web.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"D\"},{\"id\":\"sat11-rw-m1-q8\",\"number\":8,\"type\":\"text_comprehension\",\"passage\":\"The following is adapted from a 2021 essay on the composer Florence Price. After Price's death in 1953, most of her manuscripts were stored in an abandoned summer home outside St. Anne, Illinois. In 2009, renovators preparing the house for demolition discovered trunks of her scores, including two violin concertos and the Fourth Symphony, works that had never been performed publicly. The recovery has prompted orchestras to reevaluate a composer who, during her lifetime, was celebrated in Chicago but largely ignored by the East Coast musical establishment.\",\"prompt\":\"The author's main purpose in the passage is to\",\"choices\":{\"A\":\"argue that Price's Fourth Symphony deserves a place in the standard orchestral repertoire.\",\"B\":\"describe a chance discovery that has reshaped the reception of a previously overlooked composer.\",\"C\":\"contrast the musical cultures of Chicago and the East Coast in the early twentieth century.\",\"D\":\"criticize the institutions that failed to preserve Price's manuscripts during her lifetime.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m1-q9\",\"number\":9,\"type\":\"text_comprehension\",\"passage\":\"In The Federalist No. 10 (1787), James Madison addressed the problem of \\\"faction\\\"—groups of citizens united by a common interest adverse to the rights of others or to the public good. Rather than attempting to eliminate factions by suppressing liberty or enforcing uniformity of opinion, Madison argued that the remedy lay in controlling their effects. A large republic, he reasoned, would encompass so many competing interests that no single faction could easily assemble a majority, and elected representatives would tend to refine the raw preferences of constituents into more deliberate judgments.\",\"prompt\":\"Based on the text, which choice best describes Madison's view of factions?\",\"choices\":{\"A\":\"Factions are an unavoidable consequence of liberty, and their dangers are best managed through the structural features of a large republic.\",\"B\":\"Factions arise chiefly from economic inequality, which representative government is well equipped to correct.\",\"C\":\"Factions pose little threat so long as elected representatives share the moral values of their constituents.\",\"D\":\"Factions should be suppressed through legal restrictions that preserve uniformity of opinion among citizens.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m1-q10\",\"number\":10,\"type\":\"text_comprehension\",\"passage\":\"The following is adapted from Edith Wharton's 1905 novel The House of Mirth. Lily Bart, a young socialite, has just encountered Lawrence Selden at a train station.\\n\\nSelden had in fact recognized her at once, and he now advanced toward her with a certain hesitancy. It was characteristic of her to express herself so quickly, and of him to measure his words before committing them to speech. He wondered, as he looked at her, how many years it would take her to learn what it had taken him so long to unlearn: that life, at its deepest, rewarded not the swift gesture but the patient observation.\",\"prompt\":\"Which choice best describes the function of the underlined portion (\\\"He wondered...observation\\\") in the text as a whole?\",\"choices\":{\"A\":\"It reveals Selden's irritation at Lily's impulsive manner and his desire to correct her.\",\"B\":\"It suggests that Selden has come to value reflection over spontaneity and senses that Lily has not yet done so.\",\"C\":\"It implies that Selden envies Lily's quickness and regrets his own deliberate temperament.\",\"D\":\"It indicates that Selden and Lily share a temperament that both have cultivated over many years.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m1-q11\",\"number\":11,\"type\":\"text_comprehension\",\"passage\":\"Astrophysicist Wendy Freedman has long maintained that the Hubble constant—a measure of the universe's expansion rate—cannot be pinned down without first resolving systematic errors in the distance ladder used to calibrate it. In a 2023 paper, Freedman and her collaborators measured distances to nearby galaxies using the tip of the red giant branch (TRGB), a technique that bypasses Cepheid variable stars. Their value of 69.8 km/s/Mpc falls between the higher figure derived from Cepheid-based measurements and the lower figure inferred from the cosmic microwave background, a discrepancy known as the \\\"Hubble tension.\\\"\",\"prompt\":\"Based on the text, what can most reasonably be inferred about the \\\"Hubble tension\\\"?\",\"choices\":{\"A\":\"Freedman's 2023 measurement definitively resolves the Hubble tension in favor of cosmic microwave background estimates.\",\"B\":\"The Hubble tension will likely be resolved once astronomers abandon the use of Cepheid variables as distance indicators.\",\"C\":\"Freedman's intermediate value suggests that unaccounted systematic errors, rather than new physics, may underlie the discrepancy.\",\"D\":\"The Hubble tension arises primarily because the cosmic microwave background cannot be observed at sufficient resolution.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m1-q12\",\"number\":12,\"type\":\"text_comprehension\",\"passage\":\"In studies of intertemporal choice, behavioral economists have long observed that people discount future rewards more steeply than classical models predict. Offered $50 today or $55 in a week, many subjects take the $50; offered $50 in a year or $55 in a year and a week, the same subjects wait. This pattern, termed \\\"hyperbolic discounting,\\\" is difficult to reconcile with the assumption that preferences remain stable across time. Richard Thaler and others have argued that the inconsistency reflects not irrationality but a conflict between a patient \\\"planner\\\" self and an impulsive \\\"doer\\\" self, each exerting influence at different moments.\",\"prompt\":\"Which choice best describes the overall structure of the text?\",\"choices\":{\"A\":\"It identifies a puzzling behavioral pattern and then summarizes one theoretical account offered to explain it.\",\"B\":\"It presents a longstanding economic theory and then describes experimental evidence that decisively refutes it.\",\"C\":\"It compares two competing schools of thought within behavioral economics and favors one over the other.\",\"D\":\"It traces the historical development of hyperbolic discounting from its origins to its current applications.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m1-q13\",\"number\":13,\"type\":\"text_comprehension\",\"passage\":\"Art historian Patricia Rubin has argued that the workshop of Andrea del Verrocchio, active in Florence during the 1470s, functioned less as a training ground for imitators than as a laboratory in which young artists—including the young Leonardo da Vinci—tested new techniques against one another. Surviving panels from the workshop often bear the hands of multiple contributors, and Rubin contends that Verrocchio deliberately assigned passages of the same painting to different apprentices so that their divergent solutions could be studied and compared.\",\"prompt\":\"Which choice best describes the author's tone in the passage?\",\"choices\":{\"A\":\"Skeptical of Rubin's interpretation and inclined to offer an alternative.\",\"B\":\"Neutral in summarizing Rubin's argument without endorsing or rejecting it.\",\"C\":\"Admiring of Verrocchio's business acumen in running a profitable workshop.\",\"D\":\"Dismissive of earlier scholarship on Renaissance apprenticeship practices.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}]" }, { "costUsd": 0.146565, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 4 SAT Reading & Writing questions.\nType: data_analysis | Test: 11 | Unit: 1 | Start Q#: 14\n\nMANDATORY: Your JSON array MUST contain exactly 4 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m1-qN\" where N is the sequential question number starting at 14.\nFor example, the first question in this batch has id \"sat11-rw-m1-q14\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m1-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 4 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m1-q14\",\"number\":14,\"type\":\"data_analysis\",\"passage\":\"Marine biologist Hina Watanabe investigated how water temperature affects the foraging efficiency of juvenile cuttlefish (Sepia officinalis). She measured the average number of successful prey captures per hour for cuttlefish housed in tanks set to four different temperatures, recording results across a six-week observation period.\",\"visual\":{\"type\":\"table\",\"title\":\"Average Prey Captures per Hour by Water Temperature\",\"headers\":[\"Water Temperature (°C)\",\"Average Captures per Hour\",\"Standard Deviation\"],\"rows\":[[\"14\",\"3.2\",\"0.6\"],[\"18\",\"5.8\",\"0.7\"],[\"22\",\"7.1\",\"0.5\"],[\"26\",\"4.4\",\"0.9\"]]},\"prompt\":\"Which statement about the relationship between water temperature and foraging efficiency is best supported by the data?\",\"choices\":{\"A\":\"Foraging efficiency increases steadily as water temperature rises across the entire tested range.\",\"B\":\"Foraging efficiency peaks at an intermediate temperature and declines at both lower and higher temperatures.\",\"C\":\"Foraging efficiency is highest at the coldest temperature tested and decreases as the water warms.\",\"D\":\"Foraging efficiency remains essentially constant regardless of changes in water temperature.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m1-q15\",\"number\":15,\"type\":\"data_analysis\",\"passage\":\"In a behavioral economics study, researcher Diego Alarcón asked participants to choose between receiving a smaller monetary reward immediately or a larger reward after a delay. He recorded the percentage of participants who chose the delayed option at each combination of delay length and reward size, hypothesizing that participants would tolerate longer delays only when the delayed reward was substantially larger than the immediate one.\",\"visual\":{\"type\":\"table\",\"title\":\"Percentage Choosing Delayed Reward over $20 Immediate Payment\",\"headers\":[\"Delay Length\",\"Delayed Reward $25\",\"Delayed Reward $50\",\"Delayed Reward $100\"],\"rows\":[[\"1 week\",\"62%\",\"88%\",\"95%\"],[\"1 month\",\"31%\",\"64%\",\"82%\"],[\"6 months\",\"9%\",\"28%\",\"57%\"]]},\"prompt\":\"Which statement about the data is most consistent with Alarcón's hypothesis?\",\"choices\":{\"A\":\"At every delay length, a larger delayed reward produced a higher percentage of participants willing to wait.\",\"B\":\"Participants always preferred the immediate $20 reward over any delayed reward offered in the study.\",\"C\":\"The percentage of participants choosing the delayed reward depended only on the delay length, not on the reward size.\",\"D\":\"Increasing the delayed reward from $25 to $100 had no measurable effect when the delay was 6 months.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m1-q16\",\"number\":16,\"type\":\"data_analysis\",\"passage\":\"Agronomist Felipe Ortega tested how nitrogen fertilizer application affects yield in a variety of winter wheat. He applied four different rates of nitrogen to otherwise identical plots and recorded the average grain yield, measured in bushels per acre, at the end of the growing season. Ortega claimed that beyond a certain rate, additional nitrogen produces only minimal gains in yield.\",\"visual\":{\"type\":\"table\",\"title\":\"Winter Wheat Yield by Nitrogen Application Rate\",\"headers\":[\"Nitrogen Applied (lb/acre)\",\"Average Yield (bushels/acre)\"],\"rows\":[[\"40\",\"52\"],[\"80\",\"71\"],[\"120\",\"83\"],[\"160\",\"85\"]]},\"prompt\":\"Which finding from the table best supports Ortega's claim?\",\"choices\":{\"A\":\"Increasing nitrogen from 40 to 80 lb/acre raised yield by 19 bushels per acre.\",\"B\":\"Increasing nitrogen from 80 to 120 lb/acre raised yield by 12 bushels per acre.\",\"C\":\"Increasing nitrogen from 120 to 160 lb/acre raised yield by only 2 bushels per acre.\",\"D\":\"The plot receiving 40 lb/acre produced the lowest yield in the study.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m1-q17\",\"number\":17,\"type\":\"data_analysis\",\"passage\":\"Sociologist Marta Kowalski surveyed adults in four metropolitan areas about their weekly volunteer hours and their reported sense of community belonging, scored on a 0-to-10 scale. She wanted to determine whether residents who volunteered more also tended to report a stronger connection to their neighborhood, while cautioning that her data could not establish a causal direction.\",\"visual\":{\"type\":\"table\",\"title\":\"Average Volunteer Hours and Belonging Scores by City\",\"headers\":[\"City\",\"Avg. Weekly Volunteer Hours\",\"Avg. Belonging Score (0-10)\"],\"rows\":[[\"Riverton\",\"1.2\",\"4.3\"],[\"Glenford\",\"2.5\",\"5.9\"],[\"Pinehill\",\"3.8\",\"7.2\"],[\"Westmere\",\"5.1\",\"8.4\"]]},\"prompt\":\"Which conclusion is most strongly supported by Kowalski's data, given her stated caution?\",\"choices\":{\"A\":\"Volunteering more hours each week causes residents to feel a stronger sense of community belonging.\",\"B\":\"Cities with higher average volunteer hours tend to have higher average belonging scores, but the data alone cannot show that one variable causes the other.\",\"C\":\"Residents in cities with low belonging scores would increase their belonging if they were required to volunteer more.\",\"D\":\"Belonging scores and volunteer hours appear to be unrelated across the four cities surveyed.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}]" }, { "costUsd": 0.11679, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 4 SAT Reading & Writing questions.\nType: claim_support | Test: 11 | Unit: 1 | Start Q#: 18\n\nMANDATORY: Your JSON array MUST contain exactly 4 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m1-qN\" where N is the sequential question number starting at 18.\nFor example, the first question in this batch has id \"sat11-rw-m1-q18\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m1-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 4 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"Archaeologist Dr. Marisela Ortega has proposed that the inhabitants of the pre-Columbian settlement of Tiwanaku engaged in long-distance trade with coastal communities as early as 500 CE, rather than relying solely on local resources as previously assumed.\",\"prompt\":\"Which finding, if true, would most directly support Dr. Ortega's hypothesis?\",\"choices\":{\"A\":\"Stone tools found at Tiwanaku exhibit craftsmanship techniques similar to those documented at other highland sites of the same period.\",\"B\":\"Ceramic vessels excavated from Tiwanaku contain residues of marine fish and mollusks that are not native to the highland region.\",\"C\":\"Agricultural terraces near Tiwanaku show evidence of sophisticated irrigation systems that could support a large population.\",\"D\":\"Burial sites at coastal communities contain ornamental objects whose stylistic features differ markedly from those at Tiwanaku.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m1-q19\",\"number\":19,\"type\":\"claim_support\",\"passage\":\"In her 2019 study of urban songbird populations, ornithologist Priya Chandran argued that house finches living in cities have adapted their vocalizations to compensate for low-frequency traffic noise, allowing them to communicate more effectively than their rural counterparts in noisy environments.\",\"prompt\":\"Which finding from Chandran's study would most directly support her claim?\",\"choices\":{\"A\":\"Urban house finches produce songs with higher minimum frequencies than rural house finches, and these songs travel farther when ambient traffic noise is present.\",\"B\":\"Rural house finches and urban house finches exhibit nearly identical body sizes, plumage patterns, and foraging behaviors across the study sites.\",\"C\":\"Urban house finches sing more frequently during early morning hours, when traffic volume in the surrounding neighborhoods tends to be at its lowest.\",\"D\":\"Rural house finch populations have remained stable over the past decade, while urban populations have fluctuated in response to habitat changes.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m1-q20\",\"number\":20,\"type\":\"claim_support\",\"passage\":\"A literary scholar contends that in Virginia Woolf's novel Mrs. Dalloway, the recurring sound of Big Ben functions not merely as a marker of time but as a mechanism that forcibly returns characters from their private reveries to the shared public world of London.\",\"prompt\":\"Which quotation from Mrs. Dalloway would most effectively illustrate the scholar's claim?\",\"choices\":{\"A\":\"\\\"For having lived in Westminster—how many years now? over twenty,—one feels even in the midst of the traffic, or waking at night, Clarissa was positive, a particular hush.\\\"\",\"B\":\"\\\"She had a perpetual sense, as she watched the taxi cabs, of being out, out, far out to sea and alone; she always had the feeling that it was very, very dangerous to live even one day.\\\"\",\"C\":\"\\\"The leaden circles dissolved in the air. Such fools we are, she thought, crossing Victoria Street, and immediately she was back in the noise of the present, hearing the omnibuses grind.\\\"\",\"D\":\"\\\"He had left a note for her, written in pencil, her favourite, saying he would be back; but the words had slipped from her mind like water through fingers.\\\"\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m1-q21\",\"number\":21,\"type\":\"claim_support\",\"passage\":\"Behavioral economist Tomás Reyes hypothesizes that consumers are more likely to complete an online purchase when the checkout process explicitly displays the number of items remaining in stock, because perceived scarcity triggers a sense of urgency that overrides hesitation about the purchase.\",\"prompt\":\"Which finding, if true, would most directly undermine Reyes's hypothesis?\",\"choices\":{\"A\":\"Shoppers presented with \\\"only 3 left in stock\\\" messaging report feeling more anxious during checkout than those who see no such messaging.\",\"B\":\"Websites displaying stock counts generate higher total revenue per visitor than websites that do not display such counts.\",\"C\":\"Completion rates at checkout remain statistically identical whether or not stock-count messaging appears, regardless of how low the count displayed.\",\"D\":\"Consumers who abandon their carts most often cite shipping costs, rather than product availability, as their primary reason for leaving.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"}]" }, { "costUsd": 0.158865, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Reading & Writing questions.\nType: standard_english_conventions | Test: 11 | Unit: 1 | Start Q#: 22\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m1-qN\" where N is the sequential question number starting at 22.\nFor example, the first question in this batch has id \"sat11-rw-m1-q22\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m1-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m1-q22\",\"number\":22,\"type\":\"standard_english_conventions\",\"passage\":\"The extensive collection of botanical illustrations housed at the Royal Horticultural Society's Lindley Library [underlined: are] considered among the finest examples of nineteenth-century scientific artistry in Europe.\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"is\",\"C\":\"were\",\"D\":\"have been\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m1-q23\",\"number\":23,\"type\":\"standard_english_conventions\",\"passage\":\"After reviewing quarterly filings submitted by several multinational firms, the auditor concluded that the [underlined: companies reports] contained irregularities warranting further investigation by regulators.\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"company's reports\",\"C\":\"companies' reports\",\"D\":\"companies reports'\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m1-q24\",\"number\":24,\"type\":\"standard_english_conventions\",\"passage\":\"The nebula NGC 6888, commonly known as the Crescent Nebula, formed approximately 250,000 years [underlined: ago, it was] shaped by powerful stellar winds emanating from a central Wolf-Rayet star.\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"ago, being\",\"C\":\"ago, and\",\"D\":\"ago; it was\"},\"answer_format\":\"mcq\",\"correct_answer\":\"D\"},{\"id\":\"sat11-rw-m1-q25\",\"number\":25,\"type\":\"standard_english_conventions\",\"passage\":\"By the time archaeologist Howard Carter reached the inner chamber of Tutankhamun's tomb in 1923, ancient looters [underlined: had already removed] a small portion of the burial goods, though the bulk of the treasure remained undisturbed.\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"have already removed\",\"C\":\"already remove\",\"D\":\"are already removing\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m1-q26\",\"number\":26,\"type\":\"standard_english_conventions\",\"passage\":\"In her report to the city council, urban planner Yuki Tanaka recommended that the revised zoning policy emphasize transparency, [underlined: encourages] community accountability, and promote equitable access to public resources.\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"encourage\",\"C\":\"encouraging\",\"D\":\"had encouraged\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}]" }, { "costUsd": 0.08049, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 4 SAT Reading & Writing questions.\nType: transitions | Test: 11 | Unit: 1 | Start Q#: 27\n\nMANDATORY: Your JSON array MUST contain exactly 4 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m1-qN\" where N is the sequential question number starting at 27.\nFor example, the first question in this batch has id \"sat11-rw-m1-q27\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m1-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 4 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m1-q27\",\"number\":27,\"type\":\"transitions\",\"passage\":\"The Antikythera mechanism, an ancient Greek device recovered from a shipwreck in 1901, was long assumed to be a simple astronomical clock. ______ recent X-ray tomography has revealed more than thirty interlocking bronze gears capable of predicting eclipses and tracking the irregular motion of the moon, suggesting a level of mechanical sophistication previously thought impossible for the second century BCE.\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"However,\",\"B\":\"Likewise,\",\"C\":\"Consequently,\",\"D\":\"For instance,\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m1-q28\",\"number\":28,\"type\":\"transitions\",\"passage\":\"Economist Elinor Ostrom challenged the prevailing view that shared resources such as fisheries and grazing lands inevitably suffer from overuse. Through decades of fieldwork in Nepal, Kenya, and the Swiss Alps, she documented communities that had sustainably managed common pools for centuries. ______ she identified eight design principles, including clearly defined boundaries and graduated sanctions, that reliably predict which communal arrangements endure.\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"Nevertheless,\",\"B\":\"From these case studies,\",\"C\":\"In contrast,\",\"D\":\"Conversely,\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m1-q29\",\"number\":29,\"type\":\"transitions\",\"passage\":\"The painter Artemisia Gentileschi worked in a profession almost entirely closed to women in seventeenth-century Italy, and her early biographers often framed her career as a curiosity rather than an artistic achievement. Art historian Mary Garrard, writing in the 1980s, argued that this framing obscured the formal rigor of Gentileschi's compositions. ______ her influential monograph reoriented scholarly attention toward the technical choices visible in works such as Judith Slaying Holofernes.\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"Nonetheless,\",\"B\":\"Similarly,\",\"C\":\"Accordingly,\",\"D\":\"In particular,\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m1-q30\",\"number\":30,\"type\":\"transitions\",\"passage\":\"Cuttlefish possess chromatophores, pigment-filled sacs in their skin that expand and contract to produce rapid color changes. These displays are used for camouflage, mating signals, and territorial warnings. ______ cuttlefish are colorblind: their single type of photoreceptor cannot distinguish wavelengths, meaning they coordinate elaborate visual performances they themselves may not fully perceive.\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"Moreover,\",\"B\":\"Therefore,\",\"C\":\"Paradoxically,\",\"D\":\"Meanwhile,\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"}]" }, { "costUsd": 0.11544, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 3 SAT Reading & Writing questions.\nType: rhetorical_synthesis | Test: 11 | Unit: 1 | Start Q#: 31\n\nMANDATORY: Your JSON array MUST contain exactly 3 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m1-qN\" where N is the sequential question number starting at 31.\nFor example, the first question in this batch has id \"sat11-rw-m1-q31\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m1-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 3 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m1-q31\",\"number\":31,\"type\":\"rhetorical_synthesis\",\"passage\":\"While researching the lyrebird, a student has taken these notes:\\n- The superb lyrebird (Menura novaehollandiae) is native to southeastern Australia.\\n- Male lyrebirds are renowned for their vocal mimicry.\\n- A 2021 study by Anastasia Dalziell found that male lyrebirds mimic the alarm calls of multiple bird species during courtship.\\n- Dalziell concluded that this mimicry may deceive females into remaining nearby, mistaking the sounds for a genuine predator threat.\",\"prompt\":\"The student wants to present the conclusion of Dalziell's 2021 study. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"Native to southeastern Australia, the superb lyrebird (Menura novaehollandiae) is famous for the vocal mimicry of its males.\",\"B\":\"In a 2021 study, Anastasia Dalziell observed that male superb lyrebirds mimic the alarm calls of multiple bird species during courtship.\",\"C\":\"Dalziell concluded that male lyrebirds' courtship mimicry may deceive females into staying nearby by simulating a predator threat.\",\"D\":\"Male superb lyrebirds, native to southeastern Australia, are capable of producing the alarm calls of several other bird species.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m1-q32\",\"number\":32,\"type\":\"rhetorical_synthesis\",\"passage\":\"While researching microplastics in marine ecosystems, a student has taken these notes:\\n- Microplastics are plastic fragments smaller than 5 millimeters.\\n- A 2022 survey of the North Atlantic found an average of 1.2 microplastic particles per cubic meter of seawater.\\n- A 2023 survey of the Mediterranean Sea found an average of 4.7 microplastic particles per cubic meter of seawater.\\n- Both surveys used identical sampling methods.\",\"prompt\":\"The student wants to emphasize a difference between the two surveys' findings. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"Microplastics, defined as plastic fragments smaller than 5 millimeters, are now found throughout the world's oceans, including the North Atlantic and the Mediterranean.\",\"B\":\"Although both surveys used identical sampling methods, the 2023 Mediterranean survey detected 4.7 microplastic particles per cubic meter, nearly four times the 1.2 particles per cubic meter recorded in the 2022 North Atlantic survey.\",\"C\":\"Two recent surveys, conducted in 2022 and 2023, measured the concentration of microplastic particles in samples drawn from the North Atlantic and the Mediterranean.\",\"D\":\"Microplastic particles—plastic fragments smaller than 5 millimeters—were detected in both the North Atlantic in 2022 and the Mediterranean in 2023.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m1-q33\",\"number\":33,\"type\":\"rhetorical_synthesis\",\"passage\":\"While researching the painter Artemisia Gentileschi, a student has taken these notes:\\n- Artemisia Gentileschi (1593–1656) was an Italian Baroque painter.\\n- She trained in the workshop of her father, Orazio Gentileschi.\\n- In 1616, she became the first woman admitted to the Accademia delle Arti del Disegno in Florence.\\n- Her paintings frequently depicted heroic women from biblical and mythological narratives.\",\"prompt\":\"The student wants to introduce Artemisia Gentileschi to an audience unfamiliar with her. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"In 1616, she became the first woman admitted to the Accademia delle Arti del Disegno in Florence.\",\"B\":\"Trained in her father Orazio's workshop, she frequently depicted heroic women from biblical and mythological narratives.\",\"C\":\"Artemisia Gentileschi (1593–1656) was an Italian Baroque painter known for her depictions of heroic women from biblical and mythological narratives.\",\"D\":\"Artemisia Gentileschi's paintings, which often featured biblical and mythological heroines, reflected the techniques she learned in her father's workshop.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"}]" }, { "costUsd": 0.08679, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Reading & Writing questions.\nType: vocabulary_in_context | Test: 11 | Unit: 2 | Start Q#: 1\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m2-qN\" where N is the sequential question number starting at 1.\nFor example, the first question in this batch has id \"sat11-rw-m2-q1\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m2-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m2-q1\",\"number\":1,\"type\":\"vocabulary_in_context\",\"passage\":\"In behavioral economics, researchers often find that small contextual cues can ______ decisions that individuals believe are made on purely rational grounds. A series of experiments by Thaler and Sunstein demonstrated that default options on retirement forms, though seemingly trivial, shifted enrollment rates by more than forty percent.\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"dictate\",\"B\":\"inflect\",\"C\":\"fabricate\",\"D\":\"commandeer\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q2\",\"number\":2,\"type\":\"vocabulary_in_context\",\"passage\":\"Although the fresco cycle in the Brancacci Chapel is now attributed largely to Masaccio, earlier attributions were often ______: nineteenth-century connoisseurs assigned individual panels to Masolino or Filippino Lippi on the basis of stylistic hunches that modern pigment analysis has since overturned.\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"conjectural\",\"B\":\"duplicitous\",\"C\":\"indifferent\",\"D\":\"exhaustive\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m2-q3\",\"number\":3,\"type\":\"vocabulary_in_context\",\"passage\":\"The protein chaperone Hsp90 does not merely fold its client molecules; it also ______ their activity by holding them in metastable conformations until a cellular signal permits full maturation. This capacity to pause and then release function is central to how eukaryotic cells cope with environmental stress.\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"abrogates\",\"B\":\"amplifies\",\"C\":\"modulates\",\"D\":\"misconstrues\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m2-q4\",\"number\":4,\"type\":\"vocabulary_in_context\",\"passage\":\"In $\\\\textit{The Social Contract}$, Rousseau argues that legitimate political authority cannot be ______ from any natural hierarchy; rather, it must arise from the free consent of equals who agree to be bound by a general will. This claim marked a decisive break from earlier theorists who grounded sovereignty in paternal or divine precedence.\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"rescinded\",\"B\":\"absolved\",\"C\":\"withheld\",\"D\":\"derived\"},\"answer_format\":\"mcq\",\"correct_answer\":\"D\"},{\"id\":\"sat11-rw-m2-q5\",\"number\":5,\"type\":\"vocabulary_in_context\",\"passage\":\"Observations from the James Webb Space Telescope have begun to ______ long-standing models of galaxy formation: massive, well-structured galaxies appear at redshifts so high that, under prevailing theories, they should not yet have had time to assemble. Cosmologists are now revisiting assumptions about the efficiency of early star formation.\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"unsettle\",\"B\":\"annihilate\",\"C\":\"ratify\",\"D\":\"postpone\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"}]" }, { "costUsd": 0.24009, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 8 SAT Reading & Writing questions.\nType: text_comprehension | Test: 11 | Unit: 2 | Start Q#: 6\n\nMANDATORY: Your JSON array MUST contain exactly 8 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m2-qN\" where N is the sequential question number starting at 6.\nFor example, the first question in this batch has id \"sat11-rw-m2-q6\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m2-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 8 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m2-q6\",\"number\":6,\"type\":\"text_comprehension\",\"passage\":\"In her 1862 letters, Emily Dickinson confided to her mentor Thomas Wentworth Higginson that she avoided publication because she feared the 'auction' of the mind. Yet she circulated dozens of poems privately, sewing them into small hand-bound booklets she called fascicles. Scholars have argued that these fascicles constitute a form of publication in themselves, one that preserved authorial control while still inviting a select readership.\",\"prompt\":\"Which choice best states the main idea of the text?\",\"choices\":{\"A\":\"Dickinson's refusal to publish her poems reflected a broader Victorian discomfort with female authorship.\",\"B\":\"Dickinson's fascicles can be understood as an alternative mode of publication rather than a retreat from readership.\",\"C\":\"Dickinson's correspondence with Higginson was the primary vehicle through which her poetry reached the public.\",\"D\":\"Dickinson's hand-bound booklets were technically inferior to commercial print but artistically superior.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q7\",\"number\":7,\"type\":\"text_comprehension\",\"passage\":\"The following is adapted from a 2019 article in a marine biology journal. Sea otters (Enhydra lutris) are often described as a keystone species because of their outsized effect on kelp-forest ecosystems. By preying on sea urchins, otters prevent the urchins from overgrazing kelp, which in turn shelters dozens of fish and invertebrate species. Recent surveys along the central California coast, however, suggest that the relationship is more conditional than once assumed: in bays where pycnopodia sea stars—another urchin predator—have collapsed, otters alone cannot suppress urchin populations enough to protect kelp.\",\"prompt\":\"What is the main purpose of the text?\",\"choices\":{\"A\":\"To argue that sea otters have been incorrectly classified as a keystone species in California kelp forests.\",\"B\":\"To describe a predator-prey relationship that has remained stable despite recent ecological disturbances.\",\"C\":\"To qualify a widely held claim about sea otters by presenting evidence that their ecological role depends on context.\",\"D\":\"To compare the hunting behaviors of sea otters and pycnopodia sea stars in protecting kelp forests.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m2-q8\",\"number\":8,\"type\":\"text_comprehension\",\"passage\":\"The following is adapted from Willa Cather's 1918 novel My Ántonia. The narrator Jim Burden recalls his grandmother's garden. I sat down in the middle of the garden, where snakes could scarcely approach unseen, and leaned my back against a warm yellow pumpkin. There were some ground-cherry bushes growing along the furrows, full of fruit. I turned back the papery triangular sheaths that protected the berries and ate a few. All about me giant grasshoppers, twice as big as any I had ever seen, were doing acrobatic feats among the dried vines. The gophers scurried up and down the ploughed ground.\",\"prompt\":\"Which choice best describes the function of the underlined details (the grasshoppers, gophers, and pumpkin) in the passage?\",\"choices\":{\"A\":\"They suggest that the garden has fallen into a state of neglect and disorder.\",\"B\":\"They establish Jim's detailed, sensory immersion in a landscape that feels abundant and alive.\",\"C\":\"They foreshadow a later conflict between Jim and the natural world of the prairie.\",\"D\":\"They contrast Jim's childhood memories with the harsher realities of farm labor.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q9\",\"number\":9,\"type\":\"text_comprehension\",\"passage\":\"Economist Elinor Ostrom, awarded the Nobel Prize in 2009, challenged the then-dominant assumption that shared resources—fisheries, grazing lands, irrigation systems—were inevitably overexploited unless governed by either private ownership or centralized state control. Drawing on fieldwork in villages from Nepal to Maine, she documented communities that had sustainably managed common resources for centuries through locally devised rules and mutual monitoring. Her work did not dispute that commons can collapse; rather, it identified the institutional conditions under which they do not.\",\"prompt\":\"Based on the text, how did Ostrom's research most likely reshape scholarly thinking about common resources?\",\"choices\":{\"A\":\"It demonstrated that privatization is the only reliable safeguard against resource depletion.\",\"B\":\"It showed that state regulation and private ownership are equally effective across all contexts.\",\"C\":\"It revealed a third pathway to sustainability that had been largely overlooked by earlier theorists.\",\"D\":\"It established that resource collapse is primarily caused by a failure of centralized enforcement.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m2-q10\",\"number\":10,\"type\":\"text_comprehension\",\"passage\":\"The following is adapted from an essay on Baroque painting. Caravaggio's 'The Calling of Saint Matthew' (1599–1600) resolves its dramatic tension through a single diagonal shaft of light that cuts across the shadowed tavern. Where earlier Mannerist painters crowded the picture plane with writhing figures and competing gestures, Caravaggio does the opposite: he empties the scene of ornament so that the viewer's eye has nowhere to travel but along the beam of light. What might in lesser hands have become theatrical instead becomes inevitable.\",\"prompt\":\"As used in the final sentence, 'inevitable' most nearly suggests that the composition feels\",\"choices\":{\"A\":\"oppressive and overburdened by symbolic weight.\",\"B\":\"compositionally necessary rather than merely staged.\",\"C\":\"derivative of earlier Mannerist conventions.\",\"D\":\"ambiguous in its spiritual implications.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q11\",\"number\":11,\"type\":\"text_comprehension\",\"passage\":\"A 2021 study of urban foxes in London by ecologist Kevin Parnell reported that foxes living within the M25 orbital motorway had, over roughly seventy generations, developed shorter, broader snouts and smaller brain cases than their rural counterparts. Parnell argued that these changes parallel the cranial shifts associated with domestication in dogs and cats, even though no human has deliberately bred the foxes. He suggested that the sustained proximity to humans, and the reduced selective pressure for wariness, may be producing a form of self-domestication.\",\"prompt\":\"Which choice best describes the structure of the text?\",\"choices\":{\"A\":\"A long-standing scientific theory is presented and then rejected in light of new genetic evidence.\",\"B\":\"A researcher's findings are summarized, followed by the interpretive framework he proposes to explain them.\",\"C\":\"Two competing explanations for an observed phenomenon are compared and one is endorsed.\",\"D\":\"An anecdotal observation is generalized into a claim about urban wildlife across Europe.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q12\",\"number\":12,\"type\":\"text_comprehension\",\"passage\":\"Historian Natalie Zemon Davis, writing about sixteenth-century France, once remarked that the archival record is 'not a mirror but a sieve.' What reaches the historian, she argued, has already been filtered by the interests of those who kept records: notaries, tax collectors, church officials. A peasant woman's dispute over a dowry survives only because a court clerk found it worth transcribing. The historian's task, then, is less to reconstruct a vanished totality than to read attentively against the grain of the sources that happen to remain.\",\"prompt\":\"Which choice best states the central claim of the text?\",\"choices\":{\"A\":\"Sixteenth-century French archives are too incomplete to support reliable historical scholarship.\",\"B\":\"Historians should prioritize the recovery of voices belonging to peasant women over those of clerks.\",\"C\":\"Because archival sources reflect the priorities of their creators, historians must interpret them critically rather than treating them as transparent windows onto the past.\",\"D\":\"The metaphor of the sieve is more accurate than the metaphor of the mirror in describing any form of historical evidence.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m2-q13\",\"number\":13,\"type\":\"text_comprehension\",\"passage\":\"In studies of infant cognition, researchers often rely on the 'violation of expectation' paradigm: infants are shown an event that either conforms to or breaks a physical rule, such as an object passing through a solid wall. Infants who look longer at the impossible event are presumed to have expected the possible one. Critics have noted, however, that longer looking times could reflect any number of reactions—surprise, confusion, visual novelty—and that inferring specific conceptual knowledge from gaze alone may credit infants with more than the data warrant.\",\"prompt\":\"Which choice best describes the author's attitude toward the violation-of-expectation paradigm?\",\"choices\":{\"A\":\"Enthusiastic endorsement of its power to reveal innate knowledge.\",\"B\":\"Cautious skepticism about the specificity of the inferences it supports.\",\"C\":\"Dismissive rejection of its value to developmental research.\",\"D\":\"Neutral description without evaluative commentary.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}]" }, { "costUsd": 0.344415, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 4 SAT Reading & Writing questions.\nType: data_analysis | Test: 11 | Unit: 2 | Start Q#: 14\n\nMANDATORY: Your JSON array MUST contain exactly 4 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m2-qN\" where N is the sequential question number starting at 14.\nFor example, the first question in this batch has id \"sat11-rw-m2-q14\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m2-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 4 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m2-q14\",\"number\":14,\"type\":\"data_analysis\",\"passage\":\"A team of plant physiologists measured the weekly growth of Boston fern specimens (Nephrolepis exaltata) cultivated under five different light intensities, holding temperature, humidity, and nutrients constant. A student summarizing the experiment writes: \\\"Although the fern responds strongly to modest increases in light at low intensities, ______\\\"\",\"visual\":{\"type\":\"table\",\"title\":\"Boston Fern Growth Rate by Light Intensity\",\"headers\":[\"Light Intensity (μmol/m²/s)\",\"Growth Rate (mm/week)\"],\"rows\":[[\"100\",\"4.2\"],[\"200\",\"7.8\"],[\"400\",\"11.5\"],[\"800\",\"12.9\"],[\"1600\",\"13.1\"]]},\"prompt\":\"Which choice most effectively uses data from the table to complete the statement?\",\"choices\":{\"A\":\"growth continues to rise in direct proportion to light intensity as it doubles from 400 to 1600 $\\\\mu mol/m^2/s$.\",\"B\":\"further increases in intensity beyond 400 $\\\\mu mol/m^2/s$ yield progressively smaller gains in growth rate.\",\"C\":\"growth rate declines sharply once light intensity exceeds 800 $\\\\mu mol/m^2/s$.\",\"D\":\"the relationship between light intensity and weekly growth is not evident in the measurements.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q15\",\"number\":15,\"type\":\"data_analysis\",\"passage\":\"In a behavioral-economics field study conducted at a Dutch supermarket, researchers varied the number of jam flavors displayed on a tasting booth and recorded both the percentage of passing shoppers who made a purchase and the percentage of purchasers who later reported being satisfied with their selection. A student summarizing the results claims that presenting more choices consistently reduces both the likelihood that a shopper will buy and the shopper's satisfaction with what was chosen.\",\"visual\":{\"type\":\"table\",\"title\":\"Purchase and Satisfaction Rates by Number of Jam Options\",\"headers\":[\"Number of Options\",\"Shoppers Who Purchased (%)\",\"Purchasers Satisfied with Choice (%)\"],\"rows\":[[\"6\",\"30\",\"84\"],[\"12\",\"22\",\"72\"],[\"18\",\"15\",\"64\"],[\"24\",\"10\",\"55\"]]},\"prompt\":\"Which choice most effectively uses data from the table to support the student's claim?\",\"choices\":{\"A\":\"At 6 options, 84% of purchasers reported satisfaction, while at 24 options only 10% of shoppers made a purchase.\",\"B\":\"As the number of options rose from 6 to 24, both the purchase rate and the satisfaction rate increased steadily.\",\"C\":\"Satisfaction rates among purchasers remained effectively constant regardless of the number of options displayed.\",\"D\":\"As the number of options rose from 6 to 24, the purchase rate fell from 30% to 10% and the satisfaction rate fell from 84% to 55%.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"D\"},{\"id\":\"sat11-rw-m2-q16\",\"number\":16,\"type\":\"data_analysis\",\"passage\":\"Astronomers detect exoplanets primarily through two techniques: the radial velocity (RV) method, which measures the stellar wobble induced by an orbiting planet, and transit photometry (TP), which measures periodic dimming as a planet crosses its star. A recent review tabulated confirmed detections by each method across four planetary mass bins. An astronomy student writes: \\\"Although the radial velocity technique is particularly effective for identifying massive planets, transit photometry has proven far more productive for detecting smaller worlds.\\\"\",\"visual\":{\"type\":\"table\",\"title\":\"Confirmed Exoplanet Detections by Method and Mass\",\"headers\":[\"Planet Mass (Earth masses)\",\"RV Method\",\"TP Method\"],\"rows\":[[\"< 1\",\"12\",\"47\"],[\"1–10\",\"89\",\"203\"],[\"10–100\",\"156\",\"118\"],[\"> 100\",\"74\",\"39\"]]},\"prompt\":\"Which choice most effectively uses data from the table to support the student's claim?\",\"choices\":{\"A\":\"For planets below 10 Earth masses, transit photometry confirmed more than twice as many detections as radial velocity (250 vs. 101), whereas for planets above 100 Earth masses, radial velocity confirmed nearly twice as many as transit photometry (74 vs. 39).\",\"B\":\"Radial velocity detected 12 planets with masses below 1 Earth mass, while transit photometry detected 39 planets with masses above 100 Earth masses.\",\"C\":\"Transit photometry produced more confirmed detections than radial velocity in every mass range listed in the table.\",\"D\":\"The radial velocity method has failed to confirm any planets with masses below 1 Earth mass.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m2-q17\",\"number\":17,\"type\":\"data_analysis\",\"passage\":\"An economist at the OECD compared the tertiary-education wage premium—the percentage by which workers with a university degree out-earn otherwise comparable workers whose schooling ended at the secondary level—for two age cohorts (25–34 and 55–64) in four member countries. A student summarizing the findings claims that in every country surveyed the wage premium conferred by a tertiary degree grows larger over the course of a worker's career.\",\"visual\":{\"type\":\"table\",\"title\":\"Tertiary-Education Wage Premium by Country and Age Cohort (%)\",\"headers\":[\"Country\",\"Ages 25–34\",\"Ages 55–64\"],\"rows\":[[\"Germany\",\"38\",\"76\"],[\"South Korea\",\"32\",\"58\"],[\"United States\",\"57\",\"92\"],[\"Italy\",\"28\",\"49\"]]},\"prompt\":\"Which choice most effectively uses data from the table to support the student's claim?\",\"choices\":{\"A\":\"The United States displayed the highest wage premium among workers aged 55–64, at 92 percent.\",\"B\":\"The wage premium in Italy was smaller than that in Germany for both age cohorts.\",\"C\":\"In each of the four countries, the wage premium for workers aged 55–64 was substantially higher than the premium for workers aged 25–34.\",\"D\":\"South Korea showed a smaller premium than Germany for younger workers but a comparable premium for older workers.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"}]" }, { "costUsd": 0.13599, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 4 SAT Reading & Writing questions.\nType: claim_support | Test: 11 | Unit: 2 | Start Q#: 18\n\nMANDATORY: Your JSON array MUST contain exactly 4 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m2-qN\" where N is the sequential question number starting at 18.\nFor example, the first question in this batch has id \"sat11-rw-m2-q18\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m2-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 4 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m2-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A team of marine biologists studying the Pacific bluefin tuna (Thunnus orynnalis) hypothesized that the species' ability to maintain a body temperature significantly warmer than the surrounding seawater allows it to forage efficiently in colder, deeper waters where prey density is higher. To test this, the team tagged 84 individuals with biologgers that recorded muscle temperature, ambient water temperature, and depth over a twelve-month period.\",\"prompt\":\"Which finding from the study, if true, would most directly support the researchers' hypothesis?\",\"choices\":{\"A\":\"Tagged tuna spent approximately equal proportions of time in surface waters and in deeper waters, regardless of the thermal gradient between their muscle tissue and the surrounding environment.\",\"B\":\"Individuals whose muscle temperatures exceeded ambient water temperatures by the largest margins executed significantly more feeding dives into cold, prey-rich waters below 400 meters than did individuals with smaller thermal differentials.\",\"C\":\"The average muscle temperature of tagged tuna was roughly 5 degrees Celsius higher than the ambient water temperature across all depths sampled during the study period.\",\"D\":\"Tuna that remained primarily in warm surface waters exhibited higher overall growth rates than those that frequently descended into colder, deeper zones where prey were more abundant.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q19\",\"number\":19,\"type\":\"claim_support\",\"passage\":\"In her 2019 study of Florentine workshop practice, art historian Marta Ricci argues that Andrea del Verrocchio's studio in the 1470s operated less as a conventional master-apprentice hierarchy and more as a collaborative environment in which senior apprentices were entrusted with substantive portions of commissioned works, including passages that bore the master's signature.\",\"prompt\":\"Which finding, if true, would most directly support Ricci's argument?\",\"choices\":{\"A\":\"Surviving account books from Verrocchio's workshop show that the master personally received payment for all commissions completed between 1470 and 1478.\",\"B\":\"Technical analysis of Verrocchio's Baptism of Christ reveals that the angel on the left and portions of the landscape were executed by the young Leonardo da Vinci using a distinctly different oil-glaze technique.\",\"C\":\"Contemporary Florentine guild regulations required that any panel painting bearing a master's signature be certified as having passed inspection by the workshop's head artisan.\",\"D\":\"Verrocchio is documented to have taken on an unusually large number of apprentices during the 1470s, which scholars have attributed to growing demand for sculptural commissions in the city.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q20\",\"number\":20,\"type\":\"claim_support\",\"passage\":\"In his essay collection The Weight of Ordinary Things (2021), critic Daniel Oyelaran contends that the short stories of Yiyun Li resist conventional emotional payoffs: rather than building toward catharsis, her narratives accumulate small, unresolved observations that leave the reader in a state of productive unease.\",\"prompt\":\"Which quotation from a Yiyun Li story would most effectively illustrate Oyelaran's claim?\",\"choices\":{\"A\":\"\\\"When the letter finally arrived, Mei wept openly in the kitchen, and for the first time in years she felt something loosen in her chest, as though a door she had long kept shut had swung wide.\\\"\",\"B\":\"\\\"He noticed the faint stain on the tablecloth, the way his sister's hand hovered near her teacup without lifting it, the silence that was not quite a silence; he said nothing, and neither did she, and eventually the waiter came.\\\"\",\"C\":\"\\\"The villain's confession, delivered in the final pages with theatrical flourish, resolved every question the novel had raised and sent readers home satisfied that justice had at last been served.\\\"\",\"D\":\"\\\"Anger rose in him with sudden clarity, and he understood, as if for the first time, exactly what he must do and why; the path ahead shone bright before him.\\\"\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q21\",\"number\":21,\"type\":\"claim_support\",\"passage\":\"An economist studying regional labor markets claims that increases in a state's minimum wage lead to measurable improvements in worker retention at low-wage firms because the higher wage reduces the incentive for employees to search for better-paying positions elsewhere.\",\"prompt\":\"Which finding, if true, would most directly support the economist's claim?\",\"choices\":{\"A\":\"Following a state-level minimum wage increase, the average voluntary quit rate among workers at low-wage firms in that state declined by 18 percent relative to neighboring states that did not raise their minimum wage.\",\"B\":\"States that raised their minimum wage experienced a slight decline in the total number of low-wage jobs available during the year following the policy change.\",\"C\":\"Surveys indicate that workers who earn above their state's minimum wage generally report higher levels of job satisfaction than workers who earn at or near the minimum.\",\"D\":\"After the minimum wage increase took effect, consumer prices at low-wage firms in affected states rose modestly, partially offsetting the real wage gains for workers.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"}]" }, { "costUsd": 0.095265, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Reading & Writing questions.\nType: standard_english_conventions | Test: 11 | Unit: 2 | Start Q#: 22\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m2-qN\" where N is the sequential question number starting at 22.\nFor example, the first question in this batch has id \"sat11-rw-m2-q22\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m2-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m2-q22\",\"number\":22,\"type\":\"standard_english_conventions\",\"passage\":\"In 1977, marine biologists aboard the submersible Alvin discovered hydrothermal vents near the Galápagos Rift. The communities of tube worms, clams, and crabs clustered around these vents [underlined: thrives] in an environment once thought incapable of supporting life.\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"thrive\",\"C\":\"has thrived\",\"D\":\"thriving\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q23\",\"number\":23,\"type\":\"standard_english_conventions\",\"passage\":\"The Ghanaian architect David Adjaye is known for designs that engage with cultural memory. His firm's most celebrated project, the Smithsonian National Museum of African American History and [underlined: Culture, opened] in 2016 on the National Mall in Washington, D.C.\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"Culture opened,\",\"C\":\"Culture; opened\",\"D\":\"Culture opened\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m2-q24\",\"number\":24,\"type\":\"standard_english_conventions\",\"passage\":\"Paleontologist Mary Anning, who spent much of her life collecting fossils along the cliffs of Lyme Regis, identified the first complete ichthyosaur skeleton when she was only twelve years old. [underlined: Despite her remarkable contributions to geology, she] was denied membership in the Geological Society of London because of her gender.\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"Despite her remarkable contributions to geology she\",\"C\":\"Despite her remarkable contributions to geology; she\",\"D\":\"Despite her, remarkable contributions to geology, she\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m2-q25\",\"number\":25,\"type\":\"standard_english_conventions\",\"passage\":\"Economist Elinor Ostrom challenged the prevailing view that shared resources inevitably collapse under private exploitation. Drawing on fieldwork in Nepal, Kenya, and Switzerland, [underlined: communities that manage fisheries, forests, and irrigation systems successfully were documented by Ostrom].\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"fisheries, forests, and irrigation systems were documented as successfully managed by communities in Ostrom's work\",\"C\":\"Ostrom documented communities that successfully manage fisheries, forests, and irrigation systems\",\"D\":\"there was documentation by Ostrom of communities successfully managing fisheries, forests, and irrigation systems\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m2-q26\",\"number\":26,\"type\":\"standard_english_conventions\",\"passage\":\"The Voyager 1 spacecraft, launched in 1977, crossed into interstellar space in 2012. Although its instruments now return data from more than 15 billion miles away, [underlined: their] plutonium power source is expected to fail sometime after 2025.\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"it's\",\"C\":\"its\",\"D\":\"they're\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"}]" }, { "costUsd": 0.08484, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 4 SAT Reading & Writing questions.\nType: transitions | Test: 11 | Unit: 2 | Start Q#: 27\n\nMANDATORY: Your JSON array MUST contain exactly 4 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m2-qN\" where N is the sequential question number starting at 27.\nFor example, the first question in this batch has id \"sat11-rw-m2-q27\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m2-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 4 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m2-q27\",\"number\":27,\"type\":\"transitions\",\"passage\":\"The rediscovery of Sofonisba Anguissola in the late twentieth century prompted museums to reattribute several sixteenth-century portraits long credited to male contemporaries. Curators initially hesitated, citing a lack of signed works by the artist. ______ subsequent archival research uncovered letters from the Spanish court confirming her authorship of numerous royal portraits, including ones previously ascribed to Alonso Sánchez Coello.\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"Likewise,\",\"B\":\"However,\",\"C\":\"Therefore,\",\"D\":\"For instance,\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q28\",\"number\":28,\"type\":\"transitions\",\"passage\":\"Behavioral economist Sendhil Mullainathan has argued that scarcity imposes a cognitive tax, consuming mental bandwidth that would otherwise be available for long-term planning. In one study, participants reminded of financial stress performed significantly worse on reasoning tasks than those in a control group. ______ the effect disappeared when participants were given time to resolve the hypothetical shortfall before testing, suggesting that the impairment is situational rather than dispositional.\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"Consequently,\",\"B\":\"In addition,\",\"C\":\"Notably,\",\"D\":\"For example,\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m2-q29\",\"number\":29,\"type\":\"transitions\",\"passage\":\"Astronomers once assumed that the rotation rates of spiral galaxies would decline sharply with distance from the galactic center, mirroring the Keplerian falloff observed in our solar system. Vera Rubin's measurements of the Andromeda Galaxy in the 1970s revealed the opposite pattern: stars at the galaxy's outer edge orbited nearly as quickly as those near its core. ______ theorists proposed that vast halos of unseen matter must surround visible galaxies, providing the additional gravitational pull needed to explain the flattened rotation curves.\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"Nevertheless,\",\"B\":\"Similarly,\",\"C\":\"To account for this discrepancy,\",\"D\":\"In contrast,\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m2-q30\",\"number\":30,\"type\":\"transitions\",\"passage\":\"Political philosopher Hannah Arendt distinguished sharply between the private realm of the household and the public realm of political action, arguing that genuine freedom could exist only in the latter. Modern democracies, she contended, have blurred this distinction by treating citizens primarily as consumers whose welfare is managed by administrative bureaucracies. ______ Arendt maintained that reinvigorating the public sphere required reviving practices of deliberation and collective judgment that resist reduction to economic calculation.\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"Accordingly,\",\"B\":\"Conversely,\",\"C\":\"Meanwhile,\",\"D\":\"Granted,\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"}]" }, { "costUsd": 0.122865, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 3 SAT Reading & Writing questions.\nType: rhetorical_synthesis | Test: 11 | Unit: 2 | Start Q#: 31\n\nMANDATORY: Your JSON array MUST contain exactly 3 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m2-qN\" where N is the sequential question number starting at 31.\nFor example, the first question in this batch has id \"sat11-rw-m2-q31\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m2-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 3 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m2-q31\",\"number\":31,\"type\":\"rhetorical_synthesis\",\"passage\":\"While researching the acoustics of stringed instruments, a student has taken the following notes:\\n- The Stradivarius violins were crafted by Antonio Stradivari in Cremona, Italy, between 1666 and 1737.\\n- A 2017 double-blind study by Claudia Fritz tested 137 listeners in a concert hall.\\n- Listeners compared six Stradivarius violins with six newly made violins.\\n- Listeners preferred the sound of the new violins over the Stradivarius instruments.\\n- The result challenges the long-held belief that Stradivarius violins possess uniquely superior tonal qualities.\",\"prompt\":\"The student wants to emphasize the surprising outcome of the 2017 study. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"Antonio Stradivari crafted violins in Cremona, Italy, between 1666 and 1737, and his instruments have long been considered superior.\",\"B\":\"Although Stradivarius violins have long been prized for their supposedly unmatched tone, listeners in Claudia Fritz's 2017 double-blind study actually preferred the sound of newly made violins.\",\"C\":\"In a 2017 study conducted by Claudia Fritz, 137 listeners in a concert hall compared six Stradivarius violins with six newly made violins.\",\"D\":\"Stradivarius violins, crafted in Cremona between 1666 and 1737, were compared with newly made violins in a study involving 137 listeners.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q32\",\"number\":32,\"type\":\"rhetorical_synthesis\",\"passage\":\"While researching bioluminescent organisms, a student has taken the following notes:\\n- Dinoflagellates are single-celled marine plankton.\\n- Some species, such as Noctiluca scintillans, emit blue-green light when disturbed.\\n- The light is produced through a chemical reaction involving luciferin and the enzyme luciferase.\\n- Researchers hypothesize that this flash startles predators, giving the dinoflagellate time to escape.\\n- The phenomenon is often called the 'burglar alarm' defense.\",\"prompt\":\"The student wants to explain the proposed function of bioluminescence in Noctiluca scintillans. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"Noctiluca scintillans is a single-celled marine plankton that produces a blue-green light through a chemical reaction.\",\"B\":\"The light produced by Noctiluca scintillans is generated by a chemical reaction involving luciferin and luciferase.\",\"C\":\"When disturbed, Noctiluca scintillans emits a flash of light that researchers believe startles predators, allowing the dinoflagellate to escape—a mechanism known as the 'burglar alarm' defense.\",\"D\":\"Bioluminescent dinoflagellates, including Noctiluca scintillans, are single-celled organisms that emit blue-green light when disturbed by passing predators.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m2-q33\",\"number\":33,\"type\":\"rhetorical_synthesis\",\"passage\":\"While researching the Harlem Renaissance, a student has taken the following notes:\\n- Zora Neale Hurston (1891-1960) was an American author and anthropologist.\\n- Her novel Their Eyes Were Watching God was published in 1937.\\n- The novel received mixed reviews upon its release.\\n- Richard Wright criticized it for lacking a clear political agenda.\\n- The novel was largely forgotten until Alice Walker rediscovered it in the 1970s.\\n- It is now considered a foundational work of African American literature.\",\"prompt\":\"The student wants to emphasize a contrast between the novel's initial reception and its current status. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"Zora Neale Hurston's Their Eyes Were Watching God, published in 1937, received mixed reviews upon its release, though Richard Wright criticized it for lacking a clear political agenda.\",\"B\":\"Their Eyes Were Watching God, written by Zora Neale Hurston and published in 1937, was rediscovered by Alice Walker in the 1970s.\",\"C\":\"Although Alice Walker rediscovered Their Eyes Were Watching God in the 1970s, the novel had initially been criticized by Richard Wright.\",\"D\":\"Though initially met with mixed reviews and largely forgotten after its 1937 publication, Zora Neale Hurston's Their Eyes Were Watching God is now regarded as a foundational work of African American literature.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"D\"}]" } ], "results": [ { "attempts": 1, "index": 0, "output": { "costUsd": 0.084615, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Reading & Writing questions.\nType: vocabulary_in_context | Test: 11 | Unit: 1 | Start Q#: 1\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m1-qN\" where N is the sequential question number starting at 1.\nFor example, the first question in this batch has id \"sat11-rw-m1-q1\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m1-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m1-q1\",\"number\":1,\"type\":\"vocabulary_in_context\",\"passage\":\"In the decades following the publication of Keynes's General Theory, economists were reluctant to ______ the orthodoxy of classical market equilibrium outright. Instead, they tended to graft Keynesian insights onto existing models, preserving much of the older framework even as its foundational assumptions were being quietly eroded.\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"repudiate\",\"B\":\"disregard\",\"C\":\"misconstrue\",\"D\":\"lament\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m1-q2\",\"number\":2,\"type\":\"vocabulary_in_context\",\"passage\":\"Although Giorgio Vasari praised Pontormo's late frescoes for their technical daring, later critics found the figures' elongated limbs and unnatural palette deeply ______. To these observers, the Mannerist departure from High Renaissance proportion signaled not innovation but a troubling loss of classical composure.\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"prosaic\",\"B\":\"derivative\",\"C\":\"disconcerting\",\"D\":\"ostentatious\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m1-q3\",\"number\":3,\"type\":\"vocabulary_in_context\",\"passage\":\"Behavioral economists have long observed that consumers do not weigh gains and losses symmetrically; a loss of \\\\$100 tends to ______ the pleasure of a \\\\$100 gain. This asymmetry, known as loss aversion, helps explain why investors often hold losing stocks far longer than rational models predict.\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"eclipse\",\"B\":\"replicate\",\"C\":\"supplement\",\"D\":\"anticipate\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m1-q4\",\"number\":4,\"type\":\"vocabulary_in_context\",\"passage\":\"Because the faint signal from the distant quasar was nearly ______ by the glow of intervening galaxies, astronomers at the Keck Observatory employed adaptive optics to isolate its spectrum. Only after filtering out the foreground light could they confirm the object's redshift with any confidence.\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"amplified\",\"B\":\"refracted\",\"C\":\"subsumed\",\"D\":\"occluded\"},\"answer_format\":\"mcq\",\"correct_answer\":\"D\"},{\"id\":\"sat11-rw-m1-q5\",\"number\":5,\"type\":\"vocabulary_in_context\",\"passage\":\"Hannah Arendt argued that totalitarian regimes do not merely suppress dissent; they seek to ______ the very categories through which citizens distinguish truth from falsehood. Once this shared epistemic ground collapses, she warned, resistance itself becomes almost unintelligible.\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"codify\",\"B\":\"dissolve\",\"C\":\"arbitrate\",\"D\":\"venerate\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}]" }, "status": "ok" }, { "attempts": 1, "index": 1, "output": { "costUsd": 0.25419, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 8 SAT Reading & Writing questions.\nType: text_comprehension | Test: 11 | Unit: 1 | Start Q#: 6\n\nMANDATORY: Your JSON array MUST contain exactly 8 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m1-qN\" where N is the sequential question number starting at 6.\nFor example, the first question in this batch has id \"sat11-rw-m1-q6\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m1-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 8 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m1-q6\",\"number\":6,\"type\":\"text_comprehension\",\"passage\":\"The following is adapted from Willa Cather's 1918 novel My Ántonia. The narrator recalls his arrival at his grandparents' Nebraska farm as a child.\\n\\nI had the sense of coming home to myself, and of having found out what a little circle man's experience is. For the first time, perhaps, since that night, I was dissolved into something complete and great. When it came to me, it came as naturally as sleep. The wind sang a sort of clear tune through the long grass, and I lay down among the pumpkins and did not try to think of anything.\",\"prompt\":\"Which choice best describes the narrator's state of mind in the passage?\",\"choices\":{\"A\":\"He feels a profound sense of peace that quiets his impulse toward reflection.\",\"B\":\"He is disoriented by the unfamiliar landscape and seeks refuge among the pumpkins.\",\"C\":\"He is eager to explore the farm but exhausted from his long journey.\",\"D\":\"He is nostalgic for a former home he has recently left behind.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m1-q7\",\"number\":7,\"type\":\"text_comprehension\",\"passage\":\"In a 2019 study, ecologist Mariana Altrichter examined the role of white-lipped peccaries in the Gran Chaco forests of Paraguay. Altrichter found that the peccaries, which travel in herds of up to three hundred, create wallows—shallow muddy depressions—that persist long after the herds have moved on. These wallows retain water during dry seasons and become refuges for frogs, insects, and small mammals that would otherwise perish. When peccary populations collapse due to hunting, the wallows dry up within a single season, and local biodiversity declines sharply.\",\"prompt\":\"Which choice best states the main idea of the text?\",\"choices\":{\"A\":\"Peccary herds in the Gran Chaco have grown in size as hunting pressures have decreased in recent decades.\",\"B\":\"Altrichter's 2019 study was the first to document the behavior of white-lipped peccaries in Paraguay.\",\"C\":\"The disappearance of peccaries threatens the frogs and insects that share their habitat in the Gran Chaco.\",\"D\":\"By creating wallows that sustain other species, peccaries function as ecosystem engineers whose loss reverberates through the Chaco food web.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"D\"},{\"id\":\"sat11-rw-m1-q8\",\"number\":8,\"type\":\"text_comprehension\",\"passage\":\"The following is adapted from a 2021 essay on the composer Florence Price. After Price's death in 1953, most of her manuscripts were stored in an abandoned summer home outside St. Anne, Illinois. In 2009, renovators preparing the house for demolition discovered trunks of her scores, including two violin concertos and the Fourth Symphony, works that had never been performed publicly. The recovery has prompted orchestras to reevaluate a composer who, during her lifetime, was celebrated in Chicago but largely ignored by the East Coast musical establishment.\",\"prompt\":\"The author's main purpose in the passage is to\",\"choices\":{\"A\":\"argue that Price's Fourth Symphony deserves a place in the standard orchestral repertoire.\",\"B\":\"describe a chance discovery that has reshaped the reception of a previously overlooked composer.\",\"C\":\"contrast the musical cultures of Chicago and the East Coast in the early twentieth century.\",\"D\":\"criticize the institutions that failed to preserve Price's manuscripts during her lifetime.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m1-q9\",\"number\":9,\"type\":\"text_comprehension\",\"passage\":\"In The Federalist No. 10 (1787), James Madison addressed the problem of \\\"faction\\\"—groups of citizens united by a common interest adverse to the rights of others or to the public good. Rather than attempting to eliminate factions by suppressing liberty or enforcing uniformity of opinion, Madison argued that the remedy lay in controlling their effects. A large republic, he reasoned, would encompass so many competing interests that no single faction could easily assemble a majority, and elected representatives would tend to refine the raw preferences of constituents into more deliberate judgments.\",\"prompt\":\"Based on the text, which choice best describes Madison's view of factions?\",\"choices\":{\"A\":\"Factions are an unavoidable consequence of liberty, and their dangers are best managed through the structural features of a large republic.\",\"B\":\"Factions arise chiefly from economic inequality, which representative government is well equipped to correct.\",\"C\":\"Factions pose little threat so long as elected representatives share the moral values of their constituents.\",\"D\":\"Factions should be suppressed through legal restrictions that preserve uniformity of opinion among citizens.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m1-q10\",\"number\":10,\"type\":\"text_comprehension\",\"passage\":\"The following is adapted from Edith Wharton's 1905 novel The House of Mirth. Lily Bart, a young socialite, has just encountered Lawrence Selden at a train station.\\n\\nSelden had in fact recognized her at once, and he now advanced toward her with a certain hesitancy. It was characteristic of her to express herself so quickly, and of him to measure his words before committing them to speech. He wondered, as he looked at her, how many years it would take her to learn what it had taken him so long to unlearn: that life, at its deepest, rewarded not the swift gesture but the patient observation.\",\"prompt\":\"Which choice best describes the function of the underlined portion (\\\"He wondered...observation\\\") in the text as a whole?\",\"choices\":{\"A\":\"It reveals Selden's irritation at Lily's impulsive manner and his desire to correct her.\",\"B\":\"It suggests that Selden has come to value reflection over spontaneity and senses that Lily has not yet done so.\",\"C\":\"It implies that Selden envies Lily's quickness and regrets his own deliberate temperament.\",\"D\":\"It indicates that Selden and Lily share a temperament that both have cultivated over many years.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m1-q11\",\"number\":11,\"type\":\"text_comprehension\",\"passage\":\"Astrophysicist Wendy Freedman has long maintained that the Hubble constant—a measure of the universe's expansion rate—cannot be pinned down without first resolving systematic errors in the distance ladder used to calibrate it. In a 2023 paper, Freedman and her collaborators measured distances to nearby galaxies using the tip of the red giant branch (TRGB), a technique that bypasses Cepheid variable stars. Their value of 69.8 km/s/Mpc falls between the higher figure derived from Cepheid-based measurements and the lower figure inferred from the cosmic microwave background, a discrepancy known as the \\\"Hubble tension.\\\"\",\"prompt\":\"Based on the text, what can most reasonably be inferred about the \\\"Hubble tension\\\"?\",\"choices\":{\"A\":\"Freedman's 2023 measurement definitively resolves the Hubble tension in favor of cosmic microwave background estimates.\",\"B\":\"The Hubble tension will likely be resolved once astronomers abandon the use of Cepheid variables as distance indicators.\",\"C\":\"Freedman's intermediate value suggests that unaccounted systematic errors, rather than new physics, may underlie the discrepancy.\",\"D\":\"The Hubble tension arises primarily because the cosmic microwave background cannot be observed at sufficient resolution.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m1-q12\",\"number\":12,\"type\":\"text_comprehension\",\"passage\":\"In studies of intertemporal choice, behavioral economists have long observed that people discount future rewards more steeply than classical models predict. Offered $50 today or $55 in a week, many subjects take the $50; offered $50 in a year or $55 in a year and a week, the same subjects wait. This pattern, termed \\\"hyperbolic discounting,\\\" is difficult to reconcile with the assumption that preferences remain stable across time. Richard Thaler and others have argued that the inconsistency reflects not irrationality but a conflict between a patient \\\"planner\\\" self and an impulsive \\\"doer\\\" self, each exerting influence at different moments.\",\"prompt\":\"Which choice best describes the overall structure of the text?\",\"choices\":{\"A\":\"It identifies a puzzling behavioral pattern and then summarizes one theoretical account offered to explain it.\",\"B\":\"It presents a longstanding economic theory and then describes experimental evidence that decisively refutes it.\",\"C\":\"It compares two competing schools of thought within behavioral economics and favors one over the other.\",\"D\":\"It traces the historical development of hyperbolic discounting from its origins to its current applications.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m1-q13\",\"number\":13,\"type\":\"text_comprehension\",\"passage\":\"Art historian Patricia Rubin has argued that the workshop of Andrea del Verrocchio, active in Florence during the 1470s, functioned less as a training ground for imitators than as a laboratory in which young artists—including the young Leonardo da Vinci—tested new techniques against one another. Surviving panels from the workshop often bear the hands of multiple contributors, and Rubin contends that Verrocchio deliberately assigned passages of the same painting to different apprentices so that their divergent solutions could be studied and compared.\",\"prompt\":\"Which choice best describes the author's tone in the passage?\",\"choices\":{\"A\":\"Skeptical of Rubin's interpretation and inclined to offer an alternative.\",\"B\":\"Neutral in summarizing Rubin's argument without endorsing or rejecting it.\",\"C\":\"Admiring of Verrocchio's business acumen in running a profitable workshop.\",\"D\":\"Dismissive of earlier scholarship on Renaissance apprenticeship practices.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}]" }, "status": "ok" }, { "attempts": 1, "index": 2, "output": { "costUsd": 0.146565, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 4 SAT Reading & Writing questions.\nType: data_analysis | Test: 11 | Unit: 1 | Start Q#: 14\n\nMANDATORY: Your JSON array MUST contain exactly 4 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m1-qN\" where N is the sequential question number starting at 14.\nFor example, the first question in this batch has id \"sat11-rw-m1-q14\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m1-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 4 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m1-q14\",\"number\":14,\"type\":\"data_analysis\",\"passage\":\"Marine biologist Hina Watanabe investigated how water temperature affects the foraging efficiency of juvenile cuttlefish (Sepia officinalis). She measured the average number of successful prey captures per hour for cuttlefish housed in tanks set to four different temperatures, recording results across a six-week observation period.\",\"visual\":{\"type\":\"table\",\"title\":\"Average Prey Captures per Hour by Water Temperature\",\"headers\":[\"Water Temperature (°C)\",\"Average Captures per Hour\",\"Standard Deviation\"],\"rows\":[[\"14\",\"3.2\",\"0.6\"],[\"18\",\"5.8\",\"0.7\"],[\"22\",\"7.1\",\"0.5\"],[\"26\",\"4.4\",\"0.9\"]]},\"prompt\":\"Which statement about the relationship between water temperature and foraging efficiency is best supported by the data?\",\"choices\":{\"A\":\"Foraging efficiency increases steadily as water temperature rises across the entire tested range.\",\"B\":\"Foraging efficiency peaks at an intermediate temperature and declines at both lower and higher temperatures.\",\"C\":\"Foraging efficiency is highest at the coldest temperature tested and decreases as the water warms.\",\"D\":\"Foraging efficiency remains essentially constant regardless of changes in water temperature.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m1-q15\",\"number\":15,\"type\":\"data_analysis\",\"passage\":\"In a behavioral economics study, researcher Diego Alarcón asked participants to choose between receiving a smaller monetary reward immediately or a larger reward after a delay. He recorded the percentage of participants who chose the delayed option at each combination of delay length and reward size, hypothesizing that participants would tolerate longer delays only when the delayed reward was substantially larger than the immediate one.\",\"visual\":{\"type\":\"table\",\"title\":\"Percentage Choosing Delayed Reward over $20 Immediate Payment\",\"headers\":[\"Delay Length\",\"Delayed Reward $25\",\"Delayed Reward $50\",\"Delayed Reward $100\"],\"rows\":[[\"1 week\",\"62%\",\"88%\",\"95%\"],[\"1 month\",\"31%\",\"64%\",\"82%\"],[\"6 months\",\"9%\",\"28%\",\"57%\"]]},\"prompt\":\"Which statement about the data is most consistent with Alarcón's hypothesis?\",\"choices\":{\"A\":\"At every delay length, a larger delayed reward produced a higher percentage of participants willing to wait.\",\"B\":\"Participants always preferred the immediate $20 reward over any delayed reward offered in the study.\",\"C\":\"The percentage of participants choosing the delayed reward depended only on the delay length, not on the reward size.\",\"D\":\"Increasing the delayed reward from $25 to $100 had no measurable effect when the delay was 6 months.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m1-q16\",\"number\":16,\"type\":\"data_analysis\",\"passage\":\"Agronomist Felipe Ortega tested how nitrogen fertilizer application affects yield in a variety of winter wheat. He applied four different rates of nitrogen to otherwise identical plots and recorded the average grain yield, measured in bushels per acre, at the end of the growing season. Ortega claimed that beyond a certain rate, additional nitrogen produces only minimal gains in yield.\",\"visual\":{\"type\":\"table\",\"title\":\"Winter Wheat Yield by Nitrogen Application Rate\",\"headers\":[\"Nitrogen Applied (lb/acre)\",\"Average Yield (bushels/acre)\"],\"rows\":[[\"40\",\"52\"],[\"80\",\"71\"],[\"120\",\"83\"],[\"160\",\"85\"]]},\"prompt\":\"Which finding from the table best supports Ortega's claim?\",\"choices\":{\"A\":\"Increasing nitrogen from 40 to 80 lb/acre raised yield by 19 bushels per acre.\",\"B\":\"Increasing nitrogen from 80 to 120 lb/acre raised yield by 12 bushels per acre.\",\"C\":\"Increasing nitrogen from 120 to 160 lb/acre raised yield by only 2 bushels per acre.\",\"D\":\"The plot receiving 40 lb/acre produced the lowest yield in the study.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m1-q17\",\"number\":17,\"type\":\"data_analysis\",\"passage\":\"Sociologist Marta Kowalski surveyed adults in four metropolitan areas about their weekly volunteer hours and their reported sense of community belonging, scored on a 0-to-10 scale. She wanted to determine whether residents who volunteered more also tended to report a stronger connection to their neighborhood, while cautioning that her data could not establish a causal direction.\",\"visual\":{\"type\":\"table\",\"title\":\"Average Volunteer Hours and Belonging Scores by City\",\"headers\":[\"City\",\"Avg. Weekly Volunteer Hours\",\"Avg. Belonging Score (0-10)\"],\"rows\":[[\"Riverton\",\"1.2\",\"4.3\"],[\"Glenford\",\"2.5\",\"5.9\"],[\"Pinehill\",\"3.8\",\"7.2\"],[\"Westmere\",\"5.1\",\"8.4\"]]},\"prompt\":\"Which conclusion is most strongly supported by Kowalski's data, given her stated caution?\",\"choices\":{\"A\":\"Volunteering more hours each week causes residents to feel a stronger sense of community belonging.\",\"B\":\"Cities with higher average volunteer hours tend to have higher average belonging scores, but the data alone cannot show that one variable causes the other.\",\"C\":\"Residents in cities with low belonging scores would increase their belonging if they were required to volunteer more.\",\"D\":\"Belonging scores and volunteer hours appear to be unrelated across the four cities surveyed.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}]" }, "status": "ok" }, { "attempts": 1, "index": 3, "output": { "costUsd": 0.11679, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 4 SAT Reading & Writing questions.\nType: claim_support | Test: 11 | Unit: 1 | Start Q#: 18\n\nMANDATORY: Your JSON array MUST contain exactly 4 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m1-qN\" where N is the sequential question number starting at 18.\nFor example, the first question in this batch has id \"sat11-rw-m1-q18\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m1-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 4 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"Archaeologist Dr. Marisela Ortega has proposed that the inhabitants of the pre-Columbian settlement of Tiwanaku engaged in long-distance trade with coastal communities as early as 500 CE, rather than relying solely on local resources as previously assumed.\",\"prompt\":\"Which finding, if true, would most directly support Dr. Ortega's hypothesis?\",\"choices\":{\"A\":\"Stone tools found at Tiwanaku exhibit craftsmanship techniques similar to those documented at other highland sites of the same period.\",\"B\":\"Ceramic vessels excavated from Tiwanaku contain residues of marine fish and mollusks that are not native to the highland region.\",\"C\":\"Agricultural terraces near Tiwanaku show evidence of sophisticated irrigation systems that could support a large population.\",\"D\":\"Burial sites at coastal communities contain ornamental objects whose stylistic features differ markedly from those at Tiwanaku.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m1-q19\",\"number\":19,\"type\":\"claim_support\",\"passage\":\"In her 2019 study of urban songbird populations, ornithologist Priya Chandran argued that house finches living in cities have adapted their vocalizations to compensate for low-frequency traffic noise, allowing them to communicate more effectively than their rural counterparts in noisy environments.\",\"prompt\":\"Which finding from Chandran's study would most directly support her claim?\",\"choices\":{\"A\":\"Urban house finches produce songs with higher minimum frequencies than rural house finches, and these songs travel farther when ambient traffic noise is present.\",\"B\":\"Rural house finches and urban house finches exhibit nearly identical body sizes, plumage patterns, and foraging behaviors across the study sites.\",\"C\":\"Urban house finches sing more frequently during early morning hours, when traffic volume in the surrounding neighborhoods tends to be at its lowest.\",\"D\":\"Rural house finch populations have remained stable over the past decade, while urban populations have fluctuated in response to habitat changes.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m1-q20\",\"number\":20,\"type\":\"claim_support\",\"passage\":\"A literary scholar contends that in Virginia Woolf's novel Mrs. Dalloway, the recurring sound of Big Ben functions not merely as a marker of time but as a mechanism that forcibly returns characters from their private reveries to the shared public world of London.\",\"prompt\":\"Which quotation from Mrs. Dalloway would most effectively illustrate the scholar's claim?\",\"choices\":{\"A\":\"\\\"For having lived in Westminster—how many years now? over twenty,—one feels even in the midst of the traffic, or waking at night, Clarissa was positive, a particular hush.\\\"\",\"B\":\"\\\"She had a perpetual sense, as she watched the taxi cabs, of being out, out, far out to sea and alone; she always had the feeling that it was very, very dangerous to live even one day.\\\"\",\"C\":\"\\\"The leaden circles dissolved in the air. Such fools we are, she thought, crossing Victoria Street, and immediately she was back in the noise of the present, hearing the omnibuses grind.\\\"\",\"D\":\"\\\"He had left a note for her, written in pencil, her favourite, saying he would be back; but the words had slipped from her mind like water through fingers.\\\"\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m1-q21\",\"number\":21,\"type\":\"claim_support\",\"passage\":\"Behavioral economist Tomás Reyes hypothesizes that consumers are more likely to complete an online purchase when the checkout process explicitly displays the number of items remaining in stock, because perceived scarcity triggers a sense of urgency that overrides hesitation about the purchase.\",\"prompt\":\"Which finding, if true, would most directly undermine Reyes's hypothesis?\",\"choices\":{\"A\":\"Shoppers presented with \\\"only 3 left in stock\\\" messaging report feeling more anxious during checkout than those who see no such messaging.\",\"B\":\"Websites displaying stock counts generate higher total revenue per visitor than websites that do not display such counts.\",\"C\":\"Completion rates at checkout remain statistically identical whether or not stock-count messaging appears, regardless of how low the count displayed.\",\"D\":\"Consumers who abandon their carts most often cite shipping costs, rather than product availability, as their primary reason for leaving.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"}]" }, "status": "ok" }, { "attempts": 1, "index": 4, "output": { "costUsd": 0.158865, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Reading & Writing questions.\nType: standard_english_conventions | Test: 11 | Unit: 1 | Start Q#: 22\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m1-qN\" where N is the sequential question number starting at 22.\nFor example, the first question in this batch has id \"sat11-rw-m1-q22\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m1-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m1-q22\",\"number\":22,\"type\":\"standard_english_conventions\",\"passage\":\"The extensive collection of botanical illustrations housed at the Royal Horticultural Society's Lindley Library [underlined: are] considered among the finest examples of nineteenth-century scientific artistry in Europe.\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"is\",\"C\":\"were\",\"D\":\"have been\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m1-q23\",\"number\":23,\"type\":\"standard_english_conventions\",\"passage\":\"After reviewing quarterly filings submitted by several multinational firms, the auditor concluded that the [underlined: companies reports] contained irregularities warranting further investigation by regulators.\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"company's reports\",\"C\":\"companies' reports\",\"D\":\"companies reports'\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m1-q24\",\"number\":24,\"type\":\"standard_english_conventions\",\"passage\":\"The nebula NGC 6888, commonly known as the Crescent Nebula, formed approximately 250,000 years [underlined: ago, it was] shaped by powerful stellar winds emanating from a central Wolf-Rayet star.\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"ago, being\",\"C\":\"ago, and\",\"D\":\"ago; it was\"},\"answer_format\":\"mcq\",\"correct_answer\":\"D\"},{\"id\":\"sat11-rw-m1-q25\",\"number\":25,\"type\":\"standard_english_conventions\",\"passage\":\"By the time archaeologist Howard Carter reached the inner chamber of Tutankhamun's tomb in 1923, ancient looters [underlined: had already removed] a small portion of the burial goods, though the bulk of the treasure remained undisturbed.\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"have already removed\",\"C\":\"already remove\",\"D\":\"are already removing\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m1-q26\",\"number\":26,\"type\":\"standard_english_conventions\",\"passage\":\"In her report to the city council, urban planner Yuki Tanaka recommended that the revised zoning policy emphasize transparency, [underlined: encourages] community accountability, and promote equitable access to public resources.\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"encourage\",\"C\":\"encouraging\",\"D\":\"had encouraged\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}]" }, "status": "ok" }, { "attempts": 1, "index": 5, "output": { "costUsd": 0.08049, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 4 SAT Reading & Writing questions.\nType: transitions | Test: 11 | Unit: 1 | Start Q#: 27\n\nMANDATORY: Your JSON array MUST contain exactly 4 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m1-qN\" where N is the sequential question number starting at 27.\nFor example, the first question in this batch has id \"sat11-rw-m1-q27\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m1-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 4 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m1-q27\",\"number\":27,\"type\":\"transitions\",\"passage\":\"The Antikythera mechanism, an ancient Greek device recovered from a shipwreck in 1901, was long assumed to be a simple astronomical clock. ______ recent X-ray tomography has revealed more than thirty interlocking bronze gears capable of predicting eclipses and tracking the irregular motion of the moon, suggesting a level of mechanical sophistication previously thought impossible for the second century BCE.\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"However,\",\"B\":\"Likewise,\",\"C\":\"Consequently,\",\"D\":\"For instance,\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m1-q28\",\"number\":28,\"type\":\"transitions\",\"passage\":\"Economist Elinor Ostrom challenged the prevailing view that shared resources such as fisheries and grazing lands inevitably suffer from overuse. Through decades of fieldwork in Nepal, Kenya, and the Swiss Alps, she documented communities that had sustainably managed common pools for centuries. ______ she identified eight design principles, including clearly defined boundaries and graduated sanctions, that reliably predict which communal arrangements endure.\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"Nevertheless,\",\"B\":\"From these case studies,\",\"C\":\"In contrast,\",\"D\":\"Conversely,\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m1-q29\",\"number\":29,\"type\":\"transitions\",\"passage\":\"The painter Artemisia Gentileschi worked in a profession almost entirely closed to women in seventeenth-century Italy, and her early biographers often framed her career as a curiosity rather than an artistic achievement. Art historian Mary Garrard, writing in the 1980s, argued that this framing obscured the formal rigor of Gentileschi's compositions. ______ her influential monograph reoriented scholarly attention toward the technical choices visible in works such as Judith Slaying Holofernes.\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"Nonetheless,\",\"B\":\"Similarly,\",\"C\":\"Accordingly,\",\"D\":\"In particular,\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m1-q30\",\"number\":30,\"type\":\"transitions\",\"passage\":\"Cuttlefish possess chromatophores, pigment-filled sacs in their skin that expand and contract to produce rapid color changes. These displays are used for camouflage, mating signals, and territorial warnings. ______ cuttlefish are colorblind: their single type of photoreceptor cannot distinguish wavelengths, meaning they coordinate elaborate visual performances they themselves may not fully perceive.\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"Moreover,\",\"B\":\"Therefore,\",\"C\":\"Paradoxically,\",\"D\":\"Meanwhile,\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"}]" }, "status": "ok" }, { "attempts": 1, "index": 6, "output": { "costUsd": 0.11544, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 3 SAT Reading & Writing questions.\nType: rhetorical_synthesis | Test: 11 | Unit: 1 | Start Q#: 31\n\nMANDATORY: Your JSON array MUST contain exactly 3 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m1-qN\" where N is the sequential question number starting at 31.\nFor example, the first question in this batch has id \"sat11-rw-m1-q31\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m1-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 3 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m1-q31\",\"number\":31,\"type\":\"rhetorical_synthesis\",\"passage\":\"While researching the lyrebird, a student has taken these notes:\\n- The superb lyrebird (Menura novaehollandiae) is native to southeastern Australia.\\n- Male lyrebirds are renowned for their vocal mimicry.\\n- A 2021 study by Anastasia Dalziell found that male lyrebirds mimic the alarm calls of multiple bird species during courtship.\\n- Dalziell concluded that this mimicry may deceive females into remaining nearby, mistaking the sounds for a genuine predator threat.\",\"prompt\":\"The student wants to present the conclusion of Dalziell's 2021 study. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"Native to southeastern Australia, the superb lyrebird (Menura novaehollandiae) is famous for the vocal mimicry of its males.\",\"B\":\"In a 2021 study, Anastasia Dalziell observed that male superb lyrebirds mimic the alarm calls of multiple bird species during courtship.\",\"C\":\"Dalziell concluded that male lyrebirds' courtship mimicry may deceive females into staying nearby by simulating a predator threat.\",\"D\":\"Male superb lyrebirds, native to southeastern Australia, are capable of producing the alarm calls of several other bird species.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m1-q32\",\"number\":32,\"type\":\"rhetorical_synthesis\",\"passage\":\"While researching microplastics in marine ecosystems, a student has taken these notes:\\n- Microplastics are plastic fragments smaller than 5 millimeters.\\n- A 2022 survey of the North Atlantic found an average of 1.2 microplastic particles per cubic meter of seawater.\\n- A 2023 survey of the Mediterranean Sea found an average of 4.7 microplastic particles per cubic meter of seawater.\\n- Both surveys used identical sampling methods.\",\"prompt\":\"The student wants to emphasize a difference between the two surveys' findings. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"Microplastics, defined as plastic fragments smaller than 5 millimeters, are now found throughout the world's oceans, including the North Atlantic and the Mediterranean.\",\"B\":\"Although both surveys used identical sampling methods, the 2023 Mediterranean survey detected 4.7 microplastic particles per cubic meter, nearly four times the 1.2 particles per cubic meter recorded in the 2022 North Atlantic survey.\",\"C\":\"Two recent surveys, conducted in 2022 and 2023, measured the concentration of microplastic particles in samples drawn from the North Atlantic and the Mediterranean.\",\"D\":\"Microplastic particles—plastic fragments smaller than 5 millimeters—were detected in both the North Atlantic in 2022 and the Mediterranean in 2023.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m1-q33\",\"number\":33,\"type\":\"rhetorical_synthesis\",\"passage\":\"While researching the painter Artemisia Gentileschi, a student has taken these notes:\\n- Artemisia Gentileschi (1593–1656) was an Italian Baroque painter.\\n- She trained in the workshop of her father, Orazio Gentileschi.\\n- In 1616, she became the first woman admitted to the Accademia delle Arti del Disegno in Florence.\\n- Her paintings frequently depicted heroic women from biblical and mythological narratives.\",\"prompt\":\"The student wants to introduce Artemisia Gentileschi to an audience unfamiliar with her. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"In 1616, she became the first woman admitted to the Accademia delle Arti del Disegno in Florence.\",\"B\":\"Trained in her father Orazio's workshop, she frequently depicted heroic women from biblical and mythological narratives.\",\"C\":\"Artemisia Gentileschi (1593–1656) was an Italian Baroque painter known for her depictions of heroic women from biblical and mythological narratives.\",\"D\":\"Artemisia Gentileschi's paintings, which often featured biblical and mythological heroines, reflected the techniques she learned in her father's workshop.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"}]" }, "status": "ok" }, { "attempts": 1, "index": 7, "output": { "costUsd": 0.08679, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Reading & Writing questions.\nType: vocabulary_in_context | Test: 11 | Unit: 2 | Start Q#: 1\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m2-qN\" where N is the sequential question number starting at 1.\nFor example, the first question in this batch has id \"sat11-rw-m2-q1\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m2-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m2-q1\",\"number\":1,\"type\":\"vocabulary_in_context\",\"passage\":\"In behavioral economics, researchers often find that small contextual cues can ______ decisions that individuals believe are made on purely rational grounds. A series of experiments by Thaler and Sunstein demonstrated that default options on retirement forms, though seemingly trivial, shifted enrollment rates by more than forty percent.\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"dictate\",\"B\":\"inflect\",\"C\":\"fabricate\",\"D\":\"commandeer\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q2\",\"number\":2,\"type\":\"vocabulary_in_context\",\"passage\":\"Although the fresco cycle in the Brancacci Chapel is now attributed largely to Masaccio, earlier attributions were often ______: nineteenth-century connoisseurs assigned individual panels to Masolino or Filippino Lippi on the basis of stylistic hunches that modern pigment analysis has since overturned.\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"conjectural\",\"B\":\"duplicitous\",\"C\":\"indifferent\",\"D\":\"exhaustive\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m2-q3\",\"number\":3,\"type\":\"vocabulary_in_context\",\"passage\":\"The protein chaperone Hsp90 does not merely fold its client molecules; it also ______ their activity by holding them in metastable conformations until a cellular signal permits full maturation. This capacity to pause and then release function is central to how eukaryotic cells cope with environmental stress.\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"abrogates\",\"B\":\"amplifies\",\"C\":\"modulates\",\"D\":\"misconstrues\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m2-q4\",\"number\":4,\"type\":\"vocabulary_in_context\",\"passage\":\"In $\\\\textit{The Social Contract}$, Rousseau argues that legitimate political authority cannot be ______ from any natural hierarchy; rather, it must arise from the free consent of equals who agree to be bound by a general will. This claim marked a decisive break from earlier theorists who grounded sovereignty in paternal or divine precedence.\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"rescinded\",\"B\":\"absolved\",\"C\":\"withheld\",\"D\":\"derived\"},\"answer_format\":\"mcq\",\"correct_answer\":\"D\"},{\"id\":\"sat11-rw-m2-q5\",\"number\":5,\"type\":\"vocabulary_in_context\",\"passage\":\"Observations from the James Webb Space Telescope have begun to ______ long-standing models of galaxy formation: massive, well-structured galaxies appear at redshifts so high that, under prevailing theories, they should not yet have had time to assemble. Cosmologists are now revisiting assumptions about the efficiency of early star formation.\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"unsettle\",\"B\":\"annihilate\",\"C\":\"ratify\",\"D\":\"postpone\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"}]" }, "status": "ok" }, { "attempts": 1, "index": 8, "output": { "costUsd": 0.24009, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 8 SAT Reading & Writing questions.\nType: text_comprehension | Test: 11 | Unit: 2 | Start Q#: 6\n\nMANDATORY: Your JSON array MUST contain exactly 8 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m2-qN\" where N is the sequential question number starting at 6.\nFor example, the first question in this batch has id \"sat11-rw-m2-q6\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m2-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 8 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m2-q6\",\"number\":6,\"type\":\"text_comprehension\",\"passage\":\"In her 1862 letters, Emily Dickinson confided to her mentor Thomas Wentworth Higginson that she avoided publication because she feared the 'auction' of the mind. Yet she circulated dozens of poems privately, sewing them into small hand-bound booklets she called fascicles. Scholars have argued that these fascicles constitute a form of publication in themselves, one that preserved authorial control while still inviting a select readership.\",\"prompt\":\"Which choice best states the main idea of the text?\",\"choices\":{\"A\":\"Dickinson's refusal to publish her poems reflected a broader Victorian discomfort with female authorship.\",\"B\":\"Dickinson's fascicles can be understood as an alternative mode of publication rather than a retreat from readership.\",\"C\":\"Dickinson's correspondence with Higginson was the primary vehicle through which her poetry reached the public.\",\"D\":\"Dickinson's hand-bound booklets were technically inferior to commercial print but artistically superior.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q7\",\"number\":7,\"type\":\"text_comprehension\",\"passage\":\"The following is adapted from a 2019 article in a marine biology journal. Sea otters (Enhydra lutris) are often described as a keystone species because of their outsized effect on kelp-forest ecosystems. By preying on sea urchins, otters prevent the urchins from overgrazing kelp, which in turn shelters dozens of fish and invertebrate species. Recent surveys along the central California coast, however, suggest that the relationship is more conditional than once assumed: in bays where pycnopodia sea stars—another urchin predator—have collapsed, otters alone cannot suppress urchin populations enough to protect kelp.\",\"prompt\":\"What is the main purpose of the text?\",\"choices\":{\"A\":\"To argue that sea otters have been incorrectly classified as a keystone species in California kelp forests.\",\"B\":\"To describe a predator-prey relationship that has remained stable despite recent ecological disturbances.\",\"C\":\"To qualify a widely held claim about sea otters by presenting evidence that their ecological role depends on context.\",\"D\":\"To compare the hunting behaviors of sea otters and pycnopodia sea stars in protecting kelp forests.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m2-q8\",\"number\":8,\"type\":\"text_comprehension\",\"passage\":\"The following is adapted from Willa Cather's 1918 novel My Ántonia. The narrator Jim Burden recalls his grandmother's garden. I sat down in the middle of the garden, where snakes could scarcely approach unseen, and leaned my back against a warm yellow pumpkin. There were some ground-cherry bushes growing along the furrows, full of fruit. I turned back the papery triangular sheaths that protected the berries and ate a few. All about me giant grasshoppers, twice as big as any I had ever seen, were doing acrobatic feats among the dried vines. The gophers scurried up and down the ploughed ground.\",\"prompt\":\"Which choice best describes the function of the underlined details (the grasshoppers, gophers, and pumpkin) in the passage?\",\"choices\":{\"A\":\"They suggest that the garden has fallen into a state of neglect and disorder.\",\"B\":\"They establish Jim's detailed, sensory immersion in a landscape that feels abundant and alive.\",\"C\":\"They foreshadow a later conflict between Jim and the natural world of the prairie.\",\"D\":\"They contrast Jim's childhood memories with the harsher realities of farm labor.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q9\",\"number\":9,\"type\":\"text_comprehension\",\"passage\":\"Economist Elinor Ostrom, awarded the Nobel Prize in 2009, challenged the then-dominant assumption that shared resources—fisheries, grazing lands, irrigation systems—were inevitably overexploited unless governed by either private ownership or centralized state control. Drawing on fieldwork in villages from Nepal to Maine, she documented communities that had sustainably managed common resources for centuries through locally devised rules and mutual monitoring. Her work did not dispute that commons can collapse; rather, it identified the institutional conditions under which they do not.\",\"prompt\":\"Based on the text, how did Ostrom's research most likely reshape scholarly thinking about common resources?\",\"choices\":{\"A\":\"It demonstrated that privatization is the only reliable safeguard against resource depletion.\",\"B\":\"It showed that state regulation and private ownership are equally effective across all contexts.\",\"C\":\"It revealed a third pathway to sustainability that had been largely overlooked by earlier theorists.\",\"D\":\"It established that resource collapse is primarily caused by a failure of centralized enforcement.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m2-q10\",\"number\":10,\"type\":\"text_comprehension\",\"passage\":\"The following is adapted from an essay on Baroque painting. Caravaggio's 'The Calling of Saint Matthew' (1599–1600) resolves its dramatic tension through a single diagonal shaft of light that cuts across the shadowed tavern. Where earlier Mannerist painters crowded the picture plane with writhing figures and competing gestures, Caravaggio does the opposite: he empties the scene of ornament so that the viewer's eye has nowhere to travel but along the beam of light. What might in lesser hands have become theatrical instead becomes inevitable.\",\"prompt\":\"As used in the final sentence, 'inevitable' most nearly suggests that the composition feels\",\"choices\":{\"A\":\"oppressive and overburdened by symbolic weight.\",\"B\":\"compositionally necessary rather than merely staged.\",\"C\":\"derivative of earlier Mannerist conventions.\",\"D\":\"ambiguous in its spiritual implications.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q11\",\"number\":11,\"type\":\"text_comprehension\",\"passage\":\"A 2021 study of urban foxes in London by ecologist Kevin Parnell reported that foxes living within the M25 orbital motorway had, over roughly seventy generations, developed shorter, broader snouts and smaller brain cases than their rural counterparts. Parnell argued that these changes parallel the cranial shifts associated with domestication in dogs and cats, even though no human has deliberately bred the foxes. He suggested that the sustained proximity to humans, and the reduced selective pressure for wariness, may be producing a form of self-domestication.\",\"prompt\":\"Which choice best describes the structure of the text?\",\"choices\":{\"A\":\"A long-standing scientific theory is presented and then rejected in light of new genetic evidence.\",\"B\":\"A researcher's findings are summarized, followed by the interpretive framework he proposes to explain them.\",\"C\":\"Two competing explanations for an observed phenomenon are compared and one is endorsed.\",\"D\":\"An anecdotal observation is generalized into a claim about urban wildlife across Europe.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q12\",\"number\":12,\"type\":\"text_comprehension\",\"passage\":\"Historian Natalie Zemon Davis, writing about sixteenth-century France, once remarked that the archival record is 'not a mirror but a sieve.' What reaches the historian, she argued, has already been filtered by the interests of those who kept records: notaries, tax collectors, church officials. A peasant woman's dispute over a dowry survives only because a court clerk found it worth transcribing. The historian's task, then, is less to reconstruct a vanished totality than to read attentively against the grain of the sources that happen to remain.\",\"prompt\":\"Which choice best states the central claim of the text?\",\"choices\":{\"A\":\"Sixteenth-century French archives are too incomplete to support reliable historical scholarship.\",\"B\":\"Historians should prioritize the recovery of voices belonging to peasant women over those of clerks.\",\"C\":\"Because archival sources reflect the priorities of their creators, historians must interpret them critically rather than treating them as transparent windows onto the past.\",\"D\":\"The metaphor of the sieve is more accurate than the metaphor of the mirror in describing any form of historical evidence.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m2-q13\",\"number\":13,\"type\":\"text_comprehension\",\"passage\":\"In studies of infant cognition, researchers often rely on the 'violation of expectation' paradigm: infants are shown an event that either conforms to or breaks a physical rule, such as an object passing through a solid wall. Infants who look longer at the impossible event are presumed to have expected the possible one. Critics have noted, however, that longer looking times could reflect any number of reactions—surprise, confusion, visual novelty—and that inferring specific conceptual knowledge from gaze alone may credit infants with more than the data warrant.\",\"prompt\":\"Which choice best describes the author's attitude toward the violation-of-expectation paradigm?\",\"choices\":{\"A\":\"Enthusiastic endorsement of its power to reveal innate knowledge.\",\"B\":\"Cautious skepticism about the specificity of the inferences it supports.\",\"C\":\"Dismissive rejection of its value to developmental research.\",\"D\":\"Neutral description without evaluative commentary.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}]" }, "status": "ok" }, { "attempts": 1, "index": 9, "output": { "costUsd": 0.344415, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 4 SAT Reading & Writing questions.\nType: data_analysis | Test: 11 | Unit: 2 | Start Q#: 14\n\nMANDATORY: Your JSON array MUST contain exactly 4 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m2-qN\" where N is the sequential question number starting at 14.\nFor example, the first question in this batch has id \"sat11-rw-m2-q14\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m2-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 4 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m2-q14\",\"number\":14,\"type\":\"data_analysis\",\"passage\":\"A team of plant physiologists measured the weekly growth of Boston fern specimens (Nephrolepis exaltata) cultivated under five different light intensities, holding temperature, humidity, and nutrients constant. A student summarizing the experiment writes: \\\"Although the fern responds strongly to modest increases in light at low intensities, ______\\\"\",\"visual\":{\"type\":\"table\",\"title\":\"Boston Fern Growth Rate by Light Intensity\",\"headers\":[\"Light Intensity (μmol/m²/s)\",\"Growth Rate (mm/week)\"],\"rows\":[[\"100\",\"4.2\"],[\"200\",\"7.8\"],[\"400\",\"11.5\"],[\"800\",\"12.9\"],[\"1600\",\"13.1\"]]},\"prompt\":\"Which choice most effectively uses data from the table to complete the statement?\",\"choices\":{\"A\":\"growth continues to rise in direct proportion to light intensity as it doubles from 400 to 1600 $\\\\mu mol/m^2/s$.\",\"B\":\"further increases in intensity beyond 400 $\\\\mu mol/m^2/s$ yield progressively smaller gains in growth rate.\",\"C\":\"growth rate declines sharply once light intensity exceeds 800 $\\\\mu mol/m^2/s$.\",\"D\":\"the relationship between light intensity and weekly growth is not evident in the measurements.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q15\",\"number\":15,\"type\":\"data_analysis\",\"passage\":\"In a behavioral-economics field study conducted at a Dutch supermarket, researchers varied the number of jam flavors displayed on a tasting booth and recorded both the percentage of passing shoppers who made a purchase and the percentage of purchasers who later reported being satisfied with their selection. A student summarizing the results claims that presenting more choices consistently reduces both the likelihood that a shopper will buy and the shopper's satisfaction with what was chosen.\",\"visual\":{\"type\":\"table\",\"title\":\"Purchase and Satisfaction Rates by Number of Jam Options\",\"headers\":[\"Number of Options\",\"Shoppers Who Purchased (%)\",\"Purchasers Satisfied with Choice (%)\"],\"rows\":[[\"6\",\"30\",\"84\"],[\"12\",\"22\",\"72\"],[\"18\",\"15\",\"64\"],[\"24\",\"10\",\"55\"]]},\"prompt\":\"Which choice most effectively uses data from the table to support the student's claim?\",\"choices\":{\"A\":\"At 6 options, 84% of purchasers reported satisfaction, while at 24 options only 10% of shoppers made a purchase.\",\"B\":\"As the number of options rose from 6 to 24, both the purchase rate and the satisfaction rate increased steadily.\",\"C\":\"Satisfaction rates among purchasers remained effectively constant regardless of the number of options displayed.\",\"D\":\"As the number of options rose from 6 to 24, the purchase rate fell from 30% to 10% and the satisfaction rate fell from 84% to 55%.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"D\"},{\"id\":\"sat11-rw-m2-q16\",\"number\":16,\"type\":\"data_analysis\",\"passage\":\"Astronomers detect exoplanets primarily through two techniques: the radial velocity (RV) method, which measures the stellar wobble induced by an orbiting planet, and transit photometry (TP), which measures periodic dimming as a planet crosses its star. A recent review tabulated confirmed detections by each method across four planetary mass bins. An astronomy student writes: \\\"Although the radial velocity technique is particularly effective for identifying massive planets, transit photometry has proven far more productive for detecting smaller worlds.\\\"\",\"visual\":{\"type\":\"table\",\"title\":\"Confirmed Exoplanet Detections by Method and Mass\",\"headers\":[\"Planet Mass (Earth masses)\",\"RV Method\",\"TP Method\"],\"rows\":[[\"< 1\",\"12\",\"47\"],[\"1–10\",\"89\",\"203\"],[\"10–100\",\"156\",\"118\"],[\"> 100\",\"74\",\"39\"]]},\"prompt\":\"Which choice most effectively uses data from the table to support the student's claim?\",\"choices\":{\"A\":\"For planets below 10 Earth masses, transit photometry confirmed more than twice as many detections as radial velocity (250 vs. 101), whereas for planets above 100 Earth masses, radial velocity confirmed nearly twice as many as transit photometry (74 vs. 39).\",\"B\":\"Radial velocity detected 12 planets with masses below 1 Earth mass, while transit photometry detected 39 planets with masses above 100 Earth masses.\",\"C\":\"Transit photometry produced more confirmed detections than radial velocity in every mass range listed in the table.\",\"D\":\"The radial velocity method has failed to confirm any planets with masses below 1 Earth mass.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m2-q17\",\"number\":17,\"type\":\"data_analysis\",\"passage\":\"An economist at the OECD compared the tertiary-education wage premium—the percentage by which workers with a university degree out-earn otherwise comparable workers whose schooling ended at the secondary level—for two age cohorts (25–34 and 55–64) in four member countries. A student summarizing the findings claims that in every country surveyed the wage premium conferred by a tertiary degree grows larger over the course of a worker's career.\",\"visual\":{\"type\":\"table\",\"title\":\"Tertiary-Education Wage Premium by Country and Age Cohort (%)\",\"headers\":[\"Country\",\"Ages 25–34\",\"Ages 55–64\"],\"rows\":[[\"Germany\",\"38\",\"76\"],[\"South Korea\",\"32\",\"58\"],[\"United States\",\"57\",\"92\"],[\"Italy\",\"28\",\"49\"]]},\"prompt\":\"Which choice most effectively uses data from the table to support the student's claim?\",\"choices\":{\"A\":\"The United States displayed the highest wage premium among workers aged 55–64, at 92 percent.\",\"B\":\"The wage premium in Italy was smaller than that in Germany for both age cohorts.\",\"C\":\"In each of the four countries, the wage premium for workers aged 55–64 was substantially higher than the premium for workers aged 25–34.\",\"D\":\"South Korea showed a smaller premium than Germany for younger workers but a comparable premium for older workers.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"}]" }, "status": "ok" }, { "attempts": 1, "index": 10, "output": { "costUsd": 0.13599, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 4 SAT Reading & Writing questions.\nType: claim_support | Test: 11 | Unit: 2 | Start Q#: 18\n\nMANDATORY: Your JSON array MUST contain exactly 4 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m2-qN\" where N is the sequential question number starting at 18.\nFor example, the first question in this batch has id \"sat11-rw-m2-q18\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m2-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 4 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m2-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A team of marine biologists studying the Pacific bluefin tuna (Thunnus orynnalis) hypothesized that the species' ability to maintain a body temperature significantly warmer than the surrounding seawater allows it to forage efficiently in colder, deeper waters where prey density is higher. To test this, the team tagged 84 individuals with biologgers that recorded muscle temperature, ambient water temperature, and depth over a twelve-month period.\",\"prompt\":\"Which finding from the study, if true, would most directly support the researchers' hypothesis?\",\"choices\":{\"A\":\"Tagged tuna spent approximately equal proportions of time in surface waters and in deeper waters, regardless of the thermal gradient between their muscle tissue and the surrounding environment.\",\"B\":\"Individuals whose muscle temperatures exceeded ambient water temperatures by the largest margins executed significantly more feeding dives into cold, prey-rich waters below 400 meters than did individuals with smaller thermal differentials.\",\"C\":\"The average muscle temperature of tagged tuna was roughly 5 degrees Celsius higher than the ambient water temperature across all depths sampled during the study period.\",\"D\":\"Tuna that remained primarily in warm surface waters exhibited higher overall growth rates than those that frequently descended into colder, deeper zones where prey were more abundant.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q19\",\"number\":19,\"type\":\"claim_support\",\"passage\":\"In her 2019 study of Florentine workshop practice, art historian Marta Ricci argues that Andrea del Verrocchio's studio in the 1470s operated less as a conventional master-apprentice hierarchy and more as a collaborative environment in which senior apprentices were entrusted with substantive portions of commissioned works, including passages that bore the master's signature.\",\"prompt\":\"Which finding, if true, would most directly support Ricci's argument?\",\"choices\":{\"A\":\"Surviving account books from Verrocchio's workshop show that the master personally received payment for all commissions completed between 1470 and 1478.\",\"B\":\"Technical analysis of Verrocchio's Baptism of Christ reveals that the angel on the left and portions of the landscape were executed by the young Leonardo da Vinci using a distinctly different oil-glaze technique.\",\"C\":\"Contemporary Florentine guild regulations required that any panel painting bearing a master's signature be certified as having passed inspection by the workshop's head artisan.\",\"D\":\"Verrocchio is documented to have taken on an unusually large number of apprentices during the 1470s, which scholars have attributed to growing demand for sculptural commissions in the city.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q20\",\"number\":20,\"type\":\"claim_support\",\"passage\":\"In his essay collection The Weight of Ordinary Things (2021), critic Daniel Oyelaran contends that the short stories of Yiyun Li resist conventional emotional payoffs: rather than building toward catharsis, her narratives accumulate small, unresolved observations that leave the reader in a state of productive unease.\",\"prompt\":\"Which quotation from a Yiyun Li story would most effectively illustrate Oyelaran's claim?\",\"choices\":{\"A\":\"\\\"When the letter finally arrived, Mei wept openly in the kitchen, and for the first time in years she felt something loosen in her chest, as though a door she had long kept shut had swung wide.\\\"\",\"B\":\"\\\"He noticed the faint stain on the tablecloth, the way his sister's hand hovered near her teacup without lifting it, the silence that was not quite a silence; he said nothing, and neither did she, and eventually the waiter came.\\\"\",\"C\":\"\\\"The villain's confession, delivered in the final pages with theatrical flourish, resolved every question the novel had raised and sent readers home satisfied that justice had at last been served.\\\"\",\"D\":\"\\\"Anger rose in him with sudden clarity, and he understood, as if for the first time, exactly what he must do and why; the path ahead shone bright before him.\\\"\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q21\",\"number\":21,\"type\":\"claim_support\",\"passage\":\"An economist studying regional labor markets claims that increases in a state's minimum wage lead to measurable improvements in worker retention at low-wage firms because the higher wage reduces the incentive for employees to search for better-paying positions elsewhere.\",\"prompt\":\"Which finding, if true, would most directly support the economist's claim?\",\"choices\":{\"A\":\"Following a state-level minimum wage increase, the average voluntary quit rate among workers at low-wage firms in that state declined by 18 percent relative to neighboring states that did not raise their minimum wage.\",\"B\":\"States that raised their minimum wage experienced a slight decline in the total number of low-wage jobs available during the year following the policy change.\",\"C\":\"Surveys indicate that workers who earn above their state's minimum wage generally report higher levels of job satisfaction than workers who earn at or near the minimum.\",\"D\":\"After the minimum wage increase took effect, consumer prices at low-wage firms in affected states rose modestly, partially offsetting the real wage gains for workers.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"}]" }, "status": "ok" }, { "attempts": 1, "index": 11, "output": { "costUsd": 0.095265, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Reading & Writing questions.\nType: standard_english_conventions | Test: 11 | Unit: 2 | Start Q#: 22\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m2-qN\" where N is the sequential question number starting at 22.\nFor example, the first question in this batch has id \"sat11-rw-m2-q22\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m2-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m2-q22\",\"number\":22,\"type\":\"standard_english_conventions\",\"passage\":\"In 1977, marine biologists aboard the submersible Alvin discovered hydrothermal vents near the Galápagos Rift. The communities of tube worms, clams, and crabs clustered around these vents [underlined: thrives] in an environment once thought incapable of supporting life.\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"thrive\",\"C\":\"has thrived\",\"D\":\"thriving\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q23\",\"number\":23,\"type\":\"standard_english_conventions\",\"passage\":\"The Ghanaian architect David Adjaye is known for designs that engage with cultural memory. His firm's most celebrated project, the Smithsonian National Museum of African American History and [underlined: Culture, opened] in 2016 on the National Mall in Washington, D.C.\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"Culture opened,\",\"C\":\"Culture; opened\",\"D\":\"Culture opened\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m2-q24\",\"number\":24,\"type\":\"standard_english_conventions\",\"passage\":\"Paleontologist Mary Anning, who spent much of her life collecting fossils along the cliffs of Lyme Regis, identified the first complete ichthyosaur skeleton when she was only twelve years old. [underlined: Despite her remarkable contributions to geology, she] was denied membership in the Geological Society of London because of her gender.\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"Despite her remarkable contributions to geology she\",\"C\":\"Despite her remarkable contributions to geology; she\",\"D\":\"Despite her, remarkable contributions to geology, she\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m2-q25\",\"number\":25,\"type\":\"standard_english_conventions\",\"passage\":\"Economist Elinor Ostrom challenged the prevailing view that shared resources inevitably collapse under private exploitation. Drawing on fieldwork in Nepal, Kenya, and Switzerland, [underlined: communities that manage fisheries, forests, and irrigation systems successfully were documented by Ostrom].\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"fisheries, forests, and irrigation systems were documented as successfully managed by communities in Ostrom's work\",\"C\":\"Ostrom documented communities that successfully manage fisheries, forests, and irrigation systems\",\"D\":\"there was documentation by Ostrom of communities successfully managing fisheries, forests, and irrigation systems\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m2-q26\",\"number\":26,\"type\":\"standard_english_conventions\",\"passage\":\"The Voyager 1 spacecraft, launched in 1977, crossed into interstellar space in 2012. Although its instruments now return data from more than 15 billion miles away, [underlined: their] plutonium power source is expected to fail sometime after 2025.\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"it's\",\"C\":\"its\",\"D\":\"they're\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"}]" }, "status": "ok" }, { "attempts": 1, "index": 12, "output": { "costUsd": 0.08484, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 4 SAT Reading & Writing questions.\nType: transitions | Test: 11 | Unit: 2 | Start Q#: 27\n\nMANDATORY: Your JSON array MUST contain exactly 4 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m2-qN\" where N is the sequential question number starting at 27.\nFor example, the first question in this batch has id \"sat11-rw-m2-q27\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m2-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 4 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m2-q27\",\"number\":27,\"type\":\"transitions\",\"passage\":\"The rediscovery of Sofonisba Anguissola in the late twentieth century prompted museums to reattribute several sixteenth-century portraits long credited to male contemporaries. Curators initially hesitated, citing a lack of signed works by the artist. ______ subsequent archival research uncovered letters from the Spanish court confirming her authorship of numerous royal portraits, including ones previously ascribed to Alonso Sánchez Coello.\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"Likewise,\",\"B\":\"However,\",\"C\":\"Therefore,\",\"D\":\"For instance,\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q28\",\"number\":28,\"type\":\"transitions\",\"passage\":\"Behavioral economist Sendhil Mullainathan has argued that scarcity imposes a cognitive tax, consuming mental bandwidth that would otherwise be available for long-term planning. In one study, participants reminded of financial stress performed significantly worse on reasoning tasks than those in a control group. ______ the effect disappeared when participants were given time to resolve the hypothetical shortfall before testing, suggesting that the impairment is situational rather than dispositional.\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"Consequently,\",\"B\":\"In addition,\",\"C\":\"Notably,\",\"D\":\"For example,\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m2-q29\",\"number\":29,\"type\":\"transitions\",\"passage\":\"Astronomers once assumed that the rotation rates of spiral galaxies would decline sharply with distance from the galactic center, mirroring the Keplerian falloff observed in our solar system. Vera Rubin's measurements of the Andromeda Galaxy in the 1970s revealed the opposite pattern: stars at the galaxy's outer edge orbited nearly as quickly as those near its core. ______ theorists proposed that vast halos of unseen matter must surround visible galaxies, providing the additional gravitational pull needed to explain the flattened rotation curves.\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"Nevertheless,\",\"B\":\"Similarly,\",\"C\":\"To account for this discrepancy,\",\"D\":\"In contrast,\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m2-q30\",\"number\":30,\"type\":\"transitions\",\"passage\":\"Political philosopher Hannah Arendt distinguished sharply between the private realm of the household and the public realm of political action, arguing that genuine freedom could exist only in the latter. Modern democracies, she contended, have blurred this distinction by treating citizens primarily as consumers whose welfare is managed by administrative bureaucracies. ______ Arendt maintained that reinvigorating the public sphere required reviving practices of deliberation and collective judgment that resist reduction to economic calculation.\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"Accordingly,\",\"B\":\"Conversely,\",\"C\":\"Meanwhile,\",\"D\":\"Granted,\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"}]" }, "status": "ok" }, { "attempts": 1, "index": 13, "output": { "costUsd": 0.122865, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 3 SAT Reading & Writing questions.\nType: rhetorical_synthesis | Test: 11 | Unit: 2 | Start Q#: 31\n\nMANDATORY: Your JSON array MUST contain exactly 3 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m2-qN\" where N is the sequential question number starting at 31.\nFor example, the first question in this batch has id \"sat11-rw-m2-q31\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m2-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 3 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m2-q31\",\"number\":31,\"type\":\"rhetorical_synthesis\",\"passage\":\"While researching the acoustics of stringed instruments, a student has taken the following notes:\\n- The Stradivarius violins were crafted by Antonio Stradivari in Cremona, Italy, between 1666 and 1737.\\n- A 2017 double-blind study by Claudia Fritz tested 137 listeners in a concert hall.\\n- Listeners compared six Stradivarius violins with six newly made violins.\\n- Listeners preferred the sound of the new violins over the Stradivarius instruments.\\n- The result challenges the long-held belief that Stradivarius violins possess uniquely superior tonal qualities.\",\"prompt\":\"The student wants to emphasize the surprising outcome of the 2017 study. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"Antonio Stradivari crafted violins in Cremona, Italy, between 1666 and 1737, and his instruments have long been considered superior.\",\"B\":\"Although Stradivarius violins have long been prized for their supposedly unmatched tone, listeners in Claudia Fritz's 2017 double-blind study actually preferred the sound of newly made violins.\",\"C\":\"In a 2017 study conducted by Claudia Fritz, 137 listeners in a concert hall compared six Stradivarius violins with six newly made violins.\",\"D\":\"Stradivarius violins, crafted in Cremona between 1666 and 1737, were compared with newly made violins in a study involving 137 listeners.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q32\",\"number\":32,\"type\":\"rhetorical_synthesis\",\"passage\":\"While researching bioluminescent organisms, a student has taken the following notes:\\n- Dinoflagellates are single-celled marine plankton.\\n- Some species, such as Noctiluca scintillans, emit blue-green light when disturbed.\\n- The light is produced through a chemical reaction involving luciferin and the enzyme luciferase.\\n- Researchers hypothesize that this flash startles predators, giving the dinoflagellate time to escape.\\n- The phenomenon is often called the 'burglar alarm' defense.\",\"prompt\":\"The student wants to explain the proposed function of bioluminescence in Noctiluca scintillans. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"Noctiluca scintillans is a single-celled marine plankton that produces a blue-green light through a chemical reaction.\",\"B\":\"The light produced by Noctiluca scintillans is generated by a chemical reaction involving luciferin and luciferase.\",\"C\":\"When disturbed, Noctiluca scintillans emits a flash of light that researchers believe startles predators, allowing the dinoflagellate to escape—a mechanism known as the 'burglar alarm' defense.\",\"D\":\"Bioluminescent dinoflagellates, including Noctiluca scintillans, are single-celled organisms that emit blue-green light when disturbed by passing predators.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m2-q33\",\"number\":33,\"type\":\"rhetorical_synthesis\",\"passage\":\"While researching the Harlem Renaissance, a student has taken the following notes:\\n- Zora Neale Hurston (1891-1960) was an American author and anthropologist.\\n- Her novel Their Eyes Were Watching God was published in 1937.\\n- The novel received mixed reviews upon its release.\\n- Richard Wright criticized it for lacking a clear political agenda.\\n- The novel was largely forgotten until Alice Walker rediscovered it in the 1970s.\\n- It is now considered a foundational work of African American literature.\",\"prompt\":\"The student wants to emphasize a contrast between the novel's initial reception and its current status. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"Zora Neale Hurston's Their Eyes Were Watching God, published in 1937, received mixed reviews upon its release, though Richard Wright criticized it for lacking a clear political agenda.\",\"B\":\"Their Eyes Were Watching God, written by Zora Neale Hurston and published in 1937, was rediscovered by Alice Walker in the 1970s.\",\"C\":\"Although Alice Walker rediscovered Their Eyes Were Watching God in the 1970s, the novel had initially been criticized by Richard Wright.\",\"D\":\"Though initially met with mixed reviews and largely forgotten after its 1937 publication, Zora Neale Hurston's Their Eyes Were Watching God is now regarded as a foundational work of African American literature.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"D\"}]" }, "status": "ok" } ], "succeeded": 14 }, "gen-math": { "count": 10, "costUsd": 0, "failed": 0, "items": [ { "costUsd": 0, "model": "gemini-2.5-pro", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 7 SAT Math questions (of which 2 are student-produced-response and the rest are MCQ).\nType: algebra | Test number: 11 | Module: 1 | First question number: 1\n\nMANDATORY: Your JSON array MUST contain exactly 7 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-math-m1-qN\" where N is the sequential question number starting at 1.\nFor example, the first question in this batch has id \"sat11-math-m1-q1\".\n\nMCQ Schema: {\"id\":\"sat11-math-m1-qN\",\"number\":N,\"type\":\"algebra\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\nSPR Schema: {\"id\":\"sat11-math-m1-qN\",\"number\":N,\"type\":\"algebra\",\"prompt\":\"...\",\"answer_format\":\"student_produced\",\"correct_answer\":\"numeric_value\"}\nPlace SPR questions at the END of the array.\n\nFollow instructions for your type:\n\n[algebra] (MCQ + 2 SPR)\n- Topics: linear equations (1 var, 2 vars), systems of 2 linear equations, linear inequalities, absolute value\n- Mix word problems and abstract algebra. Use $...$ notation.\n- MCQ distractors: sign errors, arithmetic mistakes, setup errors, partial answers\n- SPR answers: clean integers or simple fractions\n\n[functions_analysis] (MCQ + 2 SPR)\n- Topics: function evaluation, domain/range, transformations (f(x)+k, f(x-h)), graphical analysis, composition f(g(x)), piecewise\n- At least one conceptual insight question (e.g., (p,q) and (q,p) on same graph)\n- Describe graphs in text. All computations yield clean results.\n\n[geometry_trigonometry] (MCQ + 2 SPR)\n- Topics: circle equations, Pythagorean theorem, special right triangles, coordinate geometry, trig ratios, volume/surface area, angle relationships\n- Optional \"visual\" field: {\"type\":\"coordinate_plane\",\"description\":\"...\"}\n- Radical answers OK (e.g., $5\\sqrt{13}$). Multi-step reasoning required.\n\n[statistics_probability] (MCQ + 2 SPR)\n- Topics: mean/median/mode, standard deviation (conceptual), probability (conditional from tables), data interpretation, scatter plots/line of best fit, sampling/bias\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Distractors: P(A|B) vs P(B|A) confusion, mean vs median, causal from correlational\n\n[advanced_math] (MCQ + 2 SPR)\n- Topics: quadratics (factoring, discriminant), polynomials (zeros, end behavior), exponentials (growth/decay), rational expressions, radical expressions, non-linear systems\n- Hardest questions on the test. Multi-step reasoning required. Elegant algebraic insight preferred.\n- All expressions must be factorable with integer coefficients. SPR answers are clean numbers.\n\nVERIFY EVERY ANSWER: Solve each problem step-by-step before finalizing correct_answer.\nOutput ONLY a valid JSON array of EXACTLY 7 objects. No markdown. Count them before outputting.", "text": "```json\n[\n {\n \"id\": \"sat11-math-m1-q1\",\n \"number\": 1,\n \"type\": \"algebra\",\n \"prompt\": \"A taxi service charges a $2.50 pickup fee plus $1.75 per mile. If a passenger's total fare was $20.00, how many miles did the taxi travel?\",\n \"choices\": {\n \"A\": \"8\",\n \"B\": \"10\",\n \"C\": \"11.43\",\n \"D\": \"12.86\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"B\"\n },\n {\n \"id\": \"sat11-math-m1-q2\",\n \"number\": 2,\n \"type\": \"algebra\",\n \"prompt\": \"A small theater sells adult tickets for $12 each and child tickets for $8 each. If a total of 150 tickets were sold for a total of $1,560, how many adult tickets were sold?\",\n \"choices\": {\n \"A\": \"60\",\n \"B\": \"75\",\n \"C\": \"90\",\n \"D\": \"100\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m1-q3\",\n \"number\": 3,\n \"type\": \"algebra\",\n \"prompt\": \"Which of the following inequalities is equivalent to $7 - 2x > 15$?\",\n \"choices\": {\n \"A\": \"$x < -4$\",\n \"B\": \"$x > -4$\",\n \"C\": \"$x < 4$\",\n \"D\": \"$x > 4$\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\"\n },\n {\n \"id\": \"sat11-math-m1-q4\",\n \"number\": 4,\n \"type\": \"algebra\",\n \"prompt\": \"If $|\\frac{x}{2} - 3| = 4$, what is a possible value of $x$?\",\n \"choices\": {\n \"A\": \"2\",\n \"B\": \"7\",\n \"C\": \"-1\",\n \"D\": \"14\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"D\"\n },\n {\n \"id\": \"sat11-math-m1-q5\",\n \"number\": 5,\n \"type\": \"algebra\",\n \"prompt\": \"If $3x + 2y = 18$ and $2x + 3y = 17$, what is the value of $x+y$?\",\n \"choices\": {\n \"A\": \"1\",\n \"B\": \"4\",\n \"C\": \"7\",\n \"D\": \"35\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m1-q6\",\n \"number\": 6,\n \"type\": \"algebra\",\n \"prompt\": \"A rental company charges $40 for the first day and $25 for each additional day to rent a machine. If a rental bill for a machine was $215, for how many total days was it rented?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": \"8\"\n },\n {\n \"id\": \"sat11-math-m1-q7\",\n \"number\": 7,\n \"type\": \"algebra\",\n \"prompt\": \"The inequality $|n - 7| \\le 2$ describes the range of possible values for a variable $n$. If $n$ must be an integer, what is the largest possible value of $n$?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": \"9\"\n }\n]\n```" }, { "costUsd": 0, "model": "gemini-2.5-pro", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Math questions (of which 1 are student-produced-response and the rest are MCQ).\nType: functions_analysis | Test number: 11 | Module: 1 | First question number: 8\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-math-m1-qN\" where N is the sequential question number starting at 8.\nFor example, the first question in this batch has id \"sat11-math-m1-q8\".\n\nMCQ Schema: {\"id\":\"sat11-math-m1-qN\",\"number\":N,\"type\":\"functions_analysis\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\nSPR Schema: {\"id\":\"sat11-math-m1-qN\",\"number\":N,\"type\":\"functions_analysis\",\"prompt\":\"...\",\"answer_format\":\"student_produced\",\"correct_answer\":\"numeric_value\"}\nPlace SPR questions at the END of the array.\n\nFollow instructions for your type:\n\n[algebra] (MCQ + 1 SPR)\n- Topics: linear equations (1 var, 2 vars), systems of 2 linear equations, linear inequalities, absolute value\n- Mix word problems and abstract algebra. Use $...$ notation.\n- MCQ distractors: sign errors, arithmetic mistakes, setup errors, partial answers\n- SPR answers: clean integers or simple fractions\n\n[functions_analysis] (MCQ + 1 SPR)\n- Topics: function evaluation, domain/range, transformations (f(x)+k, f(x-h)), graphical analysis, composition f(g(x)), piecewise\n- At least one conceptual insight question (e.g., (p,q) and (q,p) on same graph)\n- Describe graphs in text. All computations yield clean results.\n\n[geometry_trigonometry] (MCQ + 1 SPR)\n- Topics: circle equations, Pythagorean theorem, special right triangles, coordinate geometry, trig ratios, volume/surface area, angle relationships\n- Optional \"visual\" field: {\"type\":\"coordinate_plane\",\"description\":\"...\"}\n- Radical answers OK (e.g., $5\\sqrt{13}$). Multi-step reasoning required.\n\n[statistics_probability] (MCQ + 1 SPR)\n- Topics: mean/median/mode, standard deviation (conceptual), probability (conditional from tables), data interpretation, scatter plots/line of best fit, sampling/bias\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Distractors: P(A|B) vs P(B|A) confusion, mean vs median, causal from correlational\n\n[advanced_math] (MCQ + 1 SPR)\n- Topics: quadratics (factoring, discriminant), polynomials (zeros, end behavior), exponentials (growth/decay), rational expressions, radical expressions, non-linear systems\n- Hardest questions on the test. Multi-step reasoning required. Elegant algebraic insight preferred.\n- All expressions must be factorable with integer coefficients. SPR answers are clean numbers.\n\nVERIFY EVERY ANSWER: Solve each problem step-by-step before finalizing correct_answer.\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "```json\n[\n {\n \"id\": \"sat11-math-m1-q8\",\n \"number\": 8,\n \"type\": \"functions_analysis\",\n \"prompt\": \"For the function $f$ defined by $f(x) = 3x^2 - 5x + 1$, what is the value of $f(-2)$?\",\n \"choices\": {\n \"A\": \"3\",\n \"B\": \"-1\",\n \"C\": \"23\",\n \"D\": \"-21\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m1-q9\",\n \"number\": 9,\n \"type\": \"functions_analysis\",\n \"prompt\": \"The graph of the function $g$ in the $xy$-plane contains the point $(2, 5)$. Which of the following points must be on the graph of the function $h$ defined by $h(x) = g(x-3) + 4$?\",\n \"choices\": {\n \"A\": \"(-1, 9)\",\n \"B\": \"(5, 9)\",\n \"C\": \"(5, 1)\",\n \"D\": \"(-1, 1)\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"B\"\n },\n {\n \"id\": \"sat11-math-m1-q10\",\n \"number\": 10,\n \"type\": \"functions_analysis\",\n \"prompt\": \"Let the functions $f$ and $g$ be defined by $f(x) = x^2 + 4$ and $g(x) = \\\\sqrt{x-1}$. What is the domain of the composite function $h(x) = f(g(x))$?\",\n \"choices\": {\n \"A\": \"x \\\\ge 1\",\n \"B\": \"x \\\\ge -4\",\n \"C\": \"x \\\\le 1\",\n \"D\": \"All real numbers\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\"\n },\n {\n \"id\": \"sat11-math-m1-q11\",\n \"number\": 11,\n \"type\": \"functions_analysis\",\n \"prompt\": \"A function $f$ is defined as follows:\\n$f(x) = \\\\begin{cases} 2x + 7, & \\\\text{if } x < 3 \\\\\\\\ x^2 - 1, & \\\\text{if } x \\\\ge 3 \\\\end{cases}$\\nWhat is the value of $f(3)$?\",\n \"choices\": {\n \"A\": \"13\",\n \"B\": \"10\",\n \"C\": \"9\",\n \"D\": \"8\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"D\"\n },\n {\n \"id\": \"sat11-math-m1-q12\",\n \"number\": 12,\n \"type\": \"functions_analysis\",\n \"prompt\": \"The functions $g$ and $h$ are defined by $g(x) = 2x - 5$ and $h(x) = x^2 + 3$. What is the value of $g(h(2))$?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": \"9\"\n }\n]\n```" }, { "costUsd": 0, "model": "gemini-2.5-pro", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Math questions (of which 1 are student-produced-response and the rest are MCQ).\nType: geometry_trigonometry | Test number: 11 | Module: 1 | First question number: 13\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-math-m1-qN\" where N is the sequential question number starting at 13.\nFor example, the first question in this batch has id \"sat11-math-m1-q13\".\n\nMCQ Schema: {\"id\":\"sat11-math-m1-qN\",\"number\":N,\"type\":\"geometry_trigonometry\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\nSPR Schema: {\"id\":\"sat11-math-m1-qN\",\"number\":N,\"type\":\"geometry_trigonometry\",\"prompt\":\"...\",\"answer_format\":\"student_produced\",\"correct_answer\":\"numeric_value\"}\nPlace SPR questions at the END of the array.\n\nFollow instructions for your type:\n\n[algebra] (MCQ + 1 SPR)\n- Topics: linear equations (1 var, 2 vars), systems of 2 linear equations, linear inequalities, absolute value\n- Mix word problems and abstract algebra. Use $...$ notation.\n- MCQ distractors: sign errors, arithmetic mistakes, setup errors, partial answers\n- SPR answers: clean integers or simple fractions\n\n[functions_analysis] (MCQ + 1 SPR)\n- Topics: function evaluation, domain/range, transformations (f(x)+k, f(x-h)), graphical analysis, composition f(g(x)), piecewise\n- At least one conceptual insight question (e.g., (p,q) and (q,p) on same graph)\n- Describe graphs in text. All computations yield clean results.\n\n[geometry_trigonometry] (MCQ + 1 SPR)\n- Topics: circle equations, Pythagorean theorem, special right triangles, coordinate geometry, trig ratios, volume/surface area, angle relationships\n- Optional \"visual\" field: {\"type\":\"coordinate_plane\",\"description\":\"...\"}\n- Radical answers OK (e.g., $5\\sqrt{13}$). Multi-step reasoning required.\n\n[statistics_probability] (MCQ + 1 SPR)\n- Topics: mean/median/mode, standard deviation (conceptual), probability (conditional from tables), data interpretation, scatter plots/line of best fit, sampling/bias\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Distractors: P(A|B) vs P(B|A) confusion, mean vs median, causal from correlational\n\n[advanced_math] (MCQ + 1 SPR)\n- Topics: quadratics (factoring, discriminant), polynomials (zeros, end behavior), exponentials (growth/decay), rational expressions, radical expressions, non-linear systems\n- Hardest questions on the test. Multi-step reasoning required. Elegant algebraic insight preferred.\n- All expressions must be factorable with integer coefficients. SPR answers are clean numbers.\n\nVERIFY EVERY ANSWER: Solve each problem step-by-step before finalizing correct_answer.\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "```json\n[\n {\n \"id\": \"sat11-math-m1-q13\",\n \"number\": 13,\n \"type\": \"geometry_trigonometry\",\n \"prompt\": \"In the xy-plane, a circle has a center at $(3, -5)$ and a radius of 7. Which of the following is the equation of the circle?\",\n \"choices\": {\n \"A\": \"$(x+3)^2 + (y-5)^2 = 7$\",\n \"B\": \"$(x-3)^2 + (y+5)^2 = 49$\",\n \"C\": \"$(x-3)^2 + (y+5)^2 = 7$\",\n \"D\": \"$(x+3)^2 + (y-5)^2 = 49$\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"B\"\n },\n {\n \"id\": \"sat11-math-m1-q14\",\n \"number\": 14,\n \"type\": \"geometry_trigonometry\",\n \"prompt\": \"In right triangle ABC, the right angle is at vertex C. If the length of side AC is 8 and the length of side BC is 15, what is the value of $\\\\cos(A)$?\",\n \"choices\": {\n \"A\": \"$8/17$\",\n \"B\": \"$15/17$\",\n \"C\": \"$8/15$\",\n \"D\": \"$17/8$\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\"\n },\n {\n \"id\": \"sat11-math-m1-q15\",\n \"number\": 15,\n \"type\": \"geometry_trigonometry\",\n \"prompt\": \"Two parallel lines are intersected by a transversal line. The measure of one of the resulting interior angles is $(3x + 10)^\\\\circ$ and the measure of its corresponding consecutive interior angle is $(5x + 30)^\\\\circ$. What is the value of $x$?\",\n \"choices\": {\n \"A\": \"10\",\n \"B\": \"17.5\",\n \"C\": \"20\",\n \"D\": \"40\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m1-q16\",\n \"number\": 16,\n \"type\": \"geometry_trigonometry\",\n \"prompt\": \"In right triangle $XYZ$, the measure of angle $Y$ is $90^\\\\circ$ and the measure of angle $X$ is $60^\\\\circ$. If side $XY$ (the side adjacent to angle $X$) has a length of 5, what is the length of the hypotenuse, $XZ$?\",\n \"choices\": {\n \"A\": \"$5\\\\sqrt{2}$\",\n \"B\": \"$5\\\\sqrt{3}$\",\n \"C\": \"$10\\\\sqrt{3}/3$\",\n \"D\": \"10\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"D\"\n },\n {\n \"id\": \"sat11-math-m1-q17\",\n \"number\": 17,\n \"type\": \"geometry_trigonometry\",\n \"prompt\": \"A rectangular prism has a height of 4 cm, a width of 3 cm, and a length of 12 cm. What is the length, in cm, of the space diagonal that connects two opposite vertices of the prism?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": \"13\"\n }\n]\n```" }, { "costUsd": 0, "model": "gemini-2.5-pro", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Math questions (of which 1 are student-produced-response and the rest are MCQ).\nType: statistics_probability | Test number: 11 | Module: 1 | First question number: 18\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-math-m1-qN\" where N is the sequential question number starting at 18.\nFor example, the first question in this batch has id \"sat11-math-m1-q18\".\n\nMCQ Schema: {\"id\":\"sat11-math-m1-qN\",\"number\":N,\"type\":\"statistics_probability\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\nSPR Schema: {\"id\":\"sat11-math-m1-qN\",\"number\":N,\"type\":\"statistics_probability\",\"prompt\":\"...\",\"answer_format\":\"student_produced\",\"correct_answer\":\"numeric_value\"}\nPlace SPR questions at the END of the array.\n\nFollow instructions for your type:\n\n[algebra] (MCQ + 1 SPR)\n- Topics: linear equations (1 var, 2 vars), systems of 2 linear equations, linear inequalities, absolute value\n- Mix word problems and abstract algebra. Use $...$ notation.\n- MCQ distractors: sign errors, arithmetic mistakes, setup errors, partial answers\n- SPR answers: clean integers or simple fractions\n\n[functions_analysis] (MCQ + 1 SPR)\n- Topics: function evaluation, domain/range, transformations (f(x)+k, f(x-h)), graphical analysis, composition f(g(x)), piecewise\n- At least one conceptual insight question (e.g., (p,q) and (q,p) on same graph)\n- Describe graphs in text. All computations yield clean results.\n\n[geometry_trigonometry] (MCQ + 1 SPR)\n- Topics: circle equations, Pythagorean theorem, special right triangles, coordinate geometry, trig ratios, volume/surface area, angle relationships\n- Optional \"visual\" field: {\"type\":\"coordinate_plane\",\"description\":\"...\"}\n- Radical answers OK (e.g., $5\\sqrt{13}$). Multi-step reasoning required.\n\n[statistics_probability] (MCQ + 1 SPR)\n- Topics: mean/median/mode, standard deviation (conceptual), probability (conditional from tables), data interpretation, scatter plots/line of best fit, sampling/bias\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Distractors: P(A|B) vs P(B|A) confusion, mean vs median, causal from correlational\n\n[advanced_math] (MCQ + 1 SPR)\n- Topics: quadratics (factoring, discriminant), polynomials (zeros, end behavior), exponentials (growth/decay), rational expressions, radical expressions, non-linear systems\n- Hardest questions on the test. Multi-step reasoning required. Elegant algebraic insight preferred.\n- All expressions must be factorable with integer coefficients. SPR answers are clean numbers.\n\nVERIFY EVERY ANSWER: Solve each problem step-by-step before finalizing correct_answer.\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "```json\n[\n {\n \"id\": \"sat11-math-m1-q18\",\n \"number\": 18,\n \"type\": \"statistics_probability\",\n \"prompt\": \"A community garden manager models the daily water usage for a plot of tomato plants. The line of best fit for the model is $y = 0.25x + 1.5$, where $y$ is the amount of water used in liters and $x$ is the average daily temperature in degrees Celsius. What is the best interpretation of the slope of this line?\",\n \"choices\": {\n \"A\": \"The initial amount of water used, regardless of temperature, is 0.25 liters.\",\n \"B\": \"For every 1-degree Celsius increase in temperature, the plants require an additional 1.5 liters of water.\",\n \"C\": \"For every 1-degree Celsius increase in average daily temperature, the predicted water usage increases by 0.25 liters.\",\n \"D\": \"For every 0.25-degree Celsius increase in temperature, the water usage increases by 1 liter.\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m1-q19\",\n \"number\": 19,\n \"type\": \"statistics_probability\",\n \"visual\": {\n \"type\": \"table\",\n \"title\": \"Instrument Played by Grade Level\",\n \"headers\": [\"\", \"Strings\", \"Woodwind\", \"Total\"],\n \"rows\": [\n [\"Grade 9\", 18, 22, 40],\n [\"Grade 10\", 24, 16, 40],\n [\"Total\", 42, 38, 80]\n ]\n },\n \"prompt\": \"The table above shows the distribution of 80 students in a music program by grade level and the type of instrument they play. If a student who plays a strings instrument is selected at random, what is the probability that the student is in Grade 9?\",\n \"choices\": {\n \"A\": \"$\\\\frac{18}{80}$\",\n \"B\": \"$\\\\frac{18}{40}$\",\n \"C\": \"$\\\\frac{40}{80}$\",\n \"D\": \"$\\\\frac{18}{42}$\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"D\"\n },\n {\n \"id\": \"sat11-math-m1-q20\",\n \"number\": 20,\n \"type\": \"statistics_probability\",\n \"prompt\": \"The list shows the number of points scored by a basketball player in each of the last 7 games: 15, 21, 18, 24, 19, 35, 17. Which statement accurately compares the mean and median of these scores?\",\n \"choices\": {\n \"A\": \"The mean is greater than the median.\",\n \"B\": \"The median is greater than the mean.\",\n \"C\": \"The mean and the median are equal.\",\n \"D\": \"The relationship cannot be determined from the information given.\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\"\n },\n {\n \"id\": \"sat11-math-m1-q21\",\n \"number\": 21,\n \"type\": \"statistics_probability\",\n \"visual\": {\n \"type\": \"table\",\n \"title\": \"City Employee Departments\",\n \"headers\": [\"Department\", \"Number of Employees\"],\n \"rows\": [\n [\"Public Works\", 450],\n [\"Parks & Recreation\", 200],\n [\"Administration\", 150],\n [\"Safety\", 700]\n ]\n },\n \"prompt\": \"A researcher wants to survey employees about workplace satisfaction in the city described in the table. Which sampling method is most likely to introduce bias?\",\n \"choices\": {\n \"A\": \"Using a random number generator to select 100 employees from a full alphabetical list of all city employees.\",\n \"B\": \"Leaving survey forms at the entrance of the Safety department building and using the first 100 that are returned.\",\n \"C\": \"Randomly selecting 50 employees from the Public Works department.\",\n \"D\": \"Creating a stratified sample by selecting employees from each department in proportion to the department's size.\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"B\"\n },\n {\n \"id\": \"sat11-math-m1-q22\",\n \"number\": 22,\n \"type\": \"statistics_probability\",\n \"prompt\": \"A dataset of 14 numbers has a mean of 68. A fifteenth number is added to the dataset, and the new mean is 70. What is the value of the number that was added to the dataset?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": 98\n }\n]\n```" }, { "costUsd": 0, "model": "gemini-2.5-pro", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Math questions (of which 2 are student-produced-response and the rest are MCQ).\nType: advanced_math | Test number: 11 | Module: 1 | First question number: 23\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-math-m1-qN\" where N is the sequential question number starting at 23.\nFor example, the first question in this batch has id \"sat11-math-m1-q23\".\n\nMCQ Schema: {\"id\":\"sat11-math-m1-qN\",\"number\":N,\"type\":\"advanced_math\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\nSPR Schema: {\"id\":\"sat11-math-m1-qN\",\"number\":N,\"type\":\"advanced_math\",\"prompt\":\"...\",\"answer_format\":\"student_produced\",\"correct_answer\":\"numeric_value\"}\nPlace SPR questions at the END of the array.\n\nFollow instructions for your type:\n\n[algebra] (MCQ + 2 SPR)\n- Topics: linear equations (1 var, 2 vars), systems of 2 linear equations, linear inequalities, absolute value\n- Mix word problems and abstract algebra. Use $...$ notation.\n- MCQ distractors: sign errors, arithmetic mistakes, setup errors, partial answers\n- SPR answers: clean integers or simple fractions\n\n[functions_analysis] (MCQ + 2 SPR)\n- Topics: function evaluation, domain/range, transformations (f(x)+k, f(x-h)), graphical analysis, composition f(g(x)), piecewise\n- At least one conceptual insight question (e.g., (p,q) and (q,p) on same graph)\n- Describe graphs in text. All computations yield clean results.\n\n[geometry_trigonometry] (MCQ + 2 SPR)\n- Topics: circle equations, Pythagorean theorem, special right triangles, coordinate geometry, trig ratios, volume/surface area, angle relationships\n- Optional \"visual\" field: {\"type\":\"coordinate_plane\",\"description\":\"...\"}\n- Radical answers OK (e.g., $5\\sqrt{13}$). Multi-step reasoning required.\n\n[statistics_probability] (MCQ + 2 SPR)\n- Topics: mean/median/mode, standard deviation (conceptual), probability (conditional from tables), data interpretation, scatter plots/line of best fit, sampling/bias\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Distractors: P(A|B) vs P(B|A) confusion, mean vs median, causal from correlational\n\n[advanced_math] (MCQ + 2 SPR)\n- Topics: quadratics (factoring, discriminant), polynomials (zeros, end behavior), exponentials (growth/decay), rational expressions, radical expressions, non-linear systems\n- Hardest questions on the test. Multi-step reasoning required. Elegant algebraic insight preferred.\n- All expressions must be factorable with integer coefficients. SPR answers are clean numbers.\n\nVERIFY EVERY ANSWER: Solve each problem step-by-step before finalizing correct_answer.\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "```json\n[\n {\n \"id\": \"sat11-math-m1-q23\",\n \"number\": 23,\n \"type\": \"advanced_math\",\n \"prompt\": \"In the xy-plane, a parabola with the equation $y = x^2 + bx + c$ and a line with the equation $y = -x + k$ intersect at exactly two points, $(2, 5)$ and $(5, 2)$. What is the value of $b$?\",\n \"choices\": {\n \"A\": \"-8\",\n \"B\": \"-6\",\n \"C\": \"2\",\n \"D\": \"7\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"B\"\n },\n {\n \"id\": \"sat11-math-m1-q24\",\n \"number\": 24,\n \"type\": \"advanced_math\",\n \"prompt\": \"Which of the following expressions is equivalent to $\\\\frac{2x^2-11x+10}{x-4}$ for $x>4$?\",\n \"choices\": {\n \"A\": \"$2x-3 - \\\\frac{2}{x-4}$\",\n \"B\": \"$2x+3 + \\\\frac{22}{x-4}$\",\n \"C\": \"$2x-19 + \\\\frac{86}{x-4}$\",\n \"D\": \"$2x-3 + \\\\frac{22}{x-4}$\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\"\n },\n {\n \"id\": \"sat11-math-m1-q25\",\n \"number\": 25,\n \"type\": \"advanced_math\",\n \"prompt\": \"The function $f$ is defined by $f(x) = k(x+a)(x+b)$, where $k$, $a$, and $b$ are constants. The graph of $y=f(x)$ in the xy-plane is a parabola with x-intercepts at $-3$ and $5$. The vertex of the parabola is $(1, -32)$. What is the value of $k$?\",\n \"choices\": {\n \"A\": \"-2\",\n \"B\": \"1\",\n \"C\": \"2\",\n \"D\": \"8\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m1-q26\",\n \"number\": 26,\n \"type\": \"advanced_math\",\n \"prompt\": \"If $x > 0$ and $\\\\sqrt{3\\\\sqrt{x}} = \\\\sqrt[n]{3x}$ for all values of $x$, what is the value of $n$?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": 4\n },\n {\n \"id\": \"sat11-math-m1-q27\",\n \"number\": 27,\n \"type\": \"advanced_math\",\n \"prompt\": \"For the quadratic equation $kx^2 - 8x + 2 = 0$, the sum of the solutions is twice the product of the solutions. What is the value of $k$?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": 2\n }\n]\n```" }, { "costUsd": 0, "model": "gemini-2.5-pro", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 7 SAT Math questions (of which 2 are student-produced-response and the rest are MCQ).\nType: algebra | Test number: 11 | Module: 2 | First question number: 1\n\nMANDATORY: Your JSON array MUST contain exactly 7 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-math-m2-qN\" where N is the sequential question number starting at 1.\nFor example, the first question in this batch has id \"sat11-math-m2-q1\".\n\nMCQ Schema: {\"id\":\"sat11-math-m2-qN\",\"number\":N,\"type\":\"algebra\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\nSPR Schema: {\"id\":\"sat11-math-m2-qN\",\"number\":N,\"type\":\"algebra\",\"prompt\":\"...\",\"answer_format\":\"student_produced\",\"correct_answer\":\"numeric_value\"}\nPlace SPR questions at the END of the array.\n\nFollow instructions for your type:\n\n[algebra] (MCQ + 2 SPR)\n- Topics: linear equations (1 var, 2 vars), systems of 2 linear equations, linear inequalities, absolute value\n- Mix word problems and abstract algebra. Use $...$ notation.\n- MCQ distractors: sign errors, arithmetic mistakes, setup errors, partial answers\n- SPR answers: clean integers or simple fractions\n\n[functions_analysis] (MCQ + 2 SPR)\n- Topics: function evaluation, domain/range, transformations (f(x)+k, f(x-h)), graphical analysis, composition f(g(x)), piecewise\n- At least one conceptual insight question (e.g., (p,q) and (q,p) on same graph)\n- Describe graphs in text. All computations yield clean results.\n\n[geometry_trigonometry] (MCQ + 2 SPR)\n- Topics: circle equations, Pythagorean theorem, special right triangles, coordinate geometry, trig ratios, volume/surface area, angle relationships\n- Optional \"visual\" field: {\"type\":\"coordinate_plane\",\"description\":\"...\"}\n- Radical answers OK (e.g., $5\\sqrt{13}$). Multi-step reasoning required.\n\n[statistics_probability] (MCQ + 2 SPR)\n- Topics: mean/median/mode, standard deviation (conceptual), probability (conditional from tables), data interpretation, scatter plots/line of best fit, sampling/bias\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Distractors: P(A|B) vs P(B|A) confusion, mean vs median, causal from correlational\n\n[advanced_math] (MCQ + 2 SPR)\n- Topics: quadratics (factoring, discriminant), polynomials (zeros, end behavior), exponentials (growth/decay), rational expressions, radical expressions, non-linear systems\n- Hardest questions on the test. Multi-step reasoning required. Elegant algebraic insight preferred.\n- All expressions must be factorable with integer coefficients. SPR answers are clean numbers.\n\nVERIFY EVERY ANSWER: Solve each problem step-by-step before finalizing correct_answer.\nOutput ONLY a valid JSON array of EXACTLY 7 objects. No markdown. Count them before outputting.", "text": "```json\n[\n {\n \"id\": \"sat11-math-m2-q1\",\n \"number\": 1,\n \"type\": \"algebra\",\n \"prompt\": \"If $\\\\frac{1}{3}(6x + 15) - 4 = 9$, what is the value of $x$?\",\n \"choices\": {\n \"A\": \"-1\",\n \"B\": \"4\",\n \"C\": \"5\",\n \"D\": \"8\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"B\"\n },\n {\n \"id\": \"sat11-math-m2-q2\",\n \"number\": 2,\n \"type\": \"algebra\",\n \"prompt\": \"A gym offers two membership plans. Plan A has a one-time initiation fee of $50 and a monthly fee of $20. Plan B has no initiation fee and a monthly fee of $30. After how many months will the total cost of both plans be equal?\",\n \"choices\": {\n \"A\": \"1.5\",\n \"B\": \"2.5\",\n \"C\": \"5\",\n \"D\": \"10\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m2-q3\",\n \"number\": 3,\n \"type\": \"algebra\",\n \"prompt\": \"Which of the following is a possible value of $y$ if $-4(y - 2) > 12$?\",\n \"choices\": {\n \"A\": \"-2\",\n \"B\": \"-1\",\n \"C\": \"0\",\n \"D\": \"2\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\"\n },\n {\n \"id\": \"sat11-math-m2-q4\",\n \"number\": 4,\n \"type\": \"algebra\",\n \"prompt\": \"If $|2k - 3| = 7$, what is the sum of the two possible values for $k$?\",\n \"choices\": {\n \"A\": \"-2\",\n \"B\": \"0\",\n \"C\": \"3\",\n \"D\": \"5\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m2-q5\",\n \"number\": 5,\n \"type\": \"algebra\",\n \"prompt\": \"If $5x - 2y = 11$ and $x + 3y = 9$, what is the value of $x$?\",\n \"choices\": {\n \"A\": \"-1\",\n \"B\": \"2\",\n \"C\": \"2.5\",\n \"D\": \"3\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"D\"\n },\n {\n \"id\": \"sat11-math-m2-q6\",\n \"number\": 6,\n \"type\": \"algebra\",\n \"prompt\": \"A rental car company charges a flat daily fee plus a fee per mile driven. A customer who drove 100 miles was charged $80. Another customer drove 150 miles on the same day and was charged $95. What is the company's flat daily fee in dollars?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": 50\n },\n {\n \"id\": \"sat11-math-m2-q7\",\n \"number\": 7,\n \"type\": \"algebra\",\n \"prompt\": \"The inequality $|x - 5| \\\\leq 2$ defines a range of values for $x$. If $x$ must be an integer, what is the sum of all possible values of $x$?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": 25\n }\n]\n```" }, { "costUsd": 0, "model": "gemini-2.5-pro", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Math questions (of which 1 are student-produced-response and the rest are MCQ).\nType: functions_analysis | Test number: 11 | Module: 2 | First question number: 8\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-math-m2-qN\" where N is the sequential question number starting at 8.\nFor example, the first question in this batch has id \"sat11-math-m2-q8\".\n\nMCQ Schema: {\"id\":\"sat11-math-m2-qN\",\"number\":N,\"type\":\"functions_analysis\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\nSPR Schema: {\"id\":\"sat11-math-m2-qN\",\"number\":N,\"type\":\"functions_analysis\",\"prompt\":\"...\",\"answer_format\":\"student_produced\",\"correct_answer\":\"numeric_value\"}\nPlace SPR questions at the END of the array.\n\nFollow instructions for your type:\n\n[algebra] (MCQ + 1 SPR)\n- Topics: linear equations (1 var, 2 vars), systems of 2 linear equations, linear inequalities, absolute value\n- Mix word problems and abstract algebra. Use $...$ notation.\n- MCQ distractors: sign errors, arithmetic mistakes, setup errors, partial answers\n- SPR answers: clean integers or simple fractions\n\n[functions_analysis] (MCQ + 1 SPR)\n- Topics: function evaluation, domain/range, transformations (f(x)+k, f(x-h)), graphical analysis, composition f(g(x)), piecewise\n- At least one conceptual insight question (e.g., (p,q) and (q,p) on same graph)\n- Describe graphs in text. All computations yield clean results.\n\n[geometry_trigonometry] (MCQ + 1 SPR)\n- Topics: circle equations, Pythagorean theorem, special right triangles, coordinate geometry, trig ratios, volume/surface area, angle relationships\n- Optional \"visual\" field: {\"type\":\"coordinate_plane\",\"description\":\"...\"}\n- Radical answers OK (e.g., $5\\sqrt{13}$). Multi-step reasoning required.\n\n[statistics_probability] (MCQ + 1 SPR)\n- Topics: mean/median/mode, standard deviation (conceptual), probability (conditional from tables), data interpretation, scatter plots/line of best fit, sampling/bias\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Distractors: P(A|B) vs P(B|A) confusion, mean vs median, causal from correlational\n\n[advanced_math] (MCQ + 1 SPR)\n- Topics: quadratics (factoring, discriminant), polynomials (zeros, end behavior), exponentials (growth/decay), rational expressions, radical expressions, non-linear systems\n- Hardest questions on the test. Multi-step reasoning required. Elegant algebraic insight preferred.\n- All expressions must be factorable with integer coefficients. SPR answers are clean numbers.\n\nVERIFY EVERY ANSWER: Solve each problem step-by-step before finalizing correct_answer.\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "```json\n[\n {\n \"id\": \"sat11-math-m2-q8\",\n \"number\": 8,\n \"type\": \"functions_analysis\",\n \"prompt\": \"Let the function $h$ be defined by $h(x) = x^2 - 3x$. If $h(a+2) = 4$ for some constant $a$, what is one possible value of $a$?\",\n \"choices\": {\n \"A\": \"2\",\n \"B\": \"-1\",\n \"C\": \"0\",\n \"D\": \"4\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\"\n },\n {\n \"id\": \"sat11-math-m2-q9\",\n \"number\": 9,\n \"type\": \"functions_analysis\",\n \"prompt\": \"The graph of a function $f$ in the $xy$-plane passes through the point $(0, 3)$. A second function, $g$, is defined by the transformation $g(x) = f(x-2) + 1$. Which of the following points must be on the graph of $g$?\",\n \"choices\": {\n \"A\": \"(-2, 4)\",\n \"B\": \"(2, 4)\",\n \"C\": \"(2, 2)\",\n \"D\": \"(-2, 2)\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"B\"\n },\n {\n \"id\": \"sat11-math-m2-q10\",\n \"number\": 10,\n \"type\": \"functions_analysis\",\n \"prompt\": \"A function $f$ is defined as $f(x) = \\\\begin{cases} 2x + 7 & \\\\text{if } x \\\\le 1 \\\\\\\\ x^2 - 4 & \\\\text{if } x > 1 \\\\end{cases}$. What is the value of $f(-3) + f(3)$?\",\n \"choices\": {\n \"A\": \"14\",\n \"B\": \"10\",\n \"C\": \"6\",\n \"D\": \"4\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m2-q11\",\n \"number\": 11,\n \"type\": \"functions_analysis\",\n \"prompt\": \"The function $g$ is defined by $g(x) = \\\\frac{x+3}{\\\\sqrt{4-x}}$. For the function's output to be a real number, which of the following inequalities describes all possible values of the input $x$?\",\n \"choices\": {\n \"A\": \"$x > 4$\",\n \"B\": \"$x \\\\ge 4$\",\n \"C\": \"$x \\\\le 4$\",\n \"D\": \"$x < 4$\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"D\"\n },\n {\n \"id\": \"sat11-math-m2-q12\",\n \"number\": 12,\n \"type\": \"functions_analysis\",\n \"prompt\": \"Let the functions $f$ and $g$ be defined by $f(x) = 2x - 5$ and $g(x) = x^2 + k$, where $k$ is a constant. If $f(g(1)) = -3$, what is the value of $k$?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": \"0\"\n }\n]\n```" }, { "costUsd": 0, "model": "gemini-2.5-pro", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Math questions (of which 1 are student-produced-response and the rest are MCQ).\nType: geometry_trigonometry | Test number: 11 | Module: 2 | First question number: 13\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-math-m2-qN\" where N is the sequential question number starting at 13.\nFor example, the first question in this batch has id \"sat11-math-m2-q13\".\n\nMCQ Schema: {\"id\":\"sat11-math-m2-qN\",\"number\":N,\"type\":\"geometry_trigonometry\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\nSPR Schema: {\"id\":\"sat11-math-m2-qN\",\"number\":N,\"type\":\"geometry_trigonometry\",\"prompt\":\"...\",\"answer_format\":\"student_produced\",\"correct_answer\":\"numeric_value\"}\nPlace SPR questions at the END of the array.\n\nFollow instructions for your type:\n\n[algebra] (MCQ + 1 SPR)\n- Topics: linear equations (1 var, 2 vars), systems of 2 linear equations, linear inequalities, absolute value\n- Mix word problems and abstract algebra. Use $...$ notation.\n- MCQ distractors: sign errors, arithmetic mistakes, setup errors, partial answers\n- SPR answers: clean integers or simple fractions\n\n[functions_analysis] (MCQ + 1 SPR)\n- Topics: function evaluation, domain/range, transformations (f(x)+k, f(x-h)), graphical analysis, composition f(g(x)), piecewise\n- At least one conceptual insight question (e.g., (p,q) and (q,p) on same graph)\n- Describe graphs in text. All computations yield clean results.\n\n[geometry_trigonometry] (MCQ + 1 SPR)\n- Topics: circle equations, Pythagorean theorem, special right triangles, coordinate geometry, trig ratios, volume/surface area, angle relationships\n- Optional \"visual\" field: {\"type\":\"coordinate_plane\",\"description\":\"...\"}\n- Radical answers OK (e.g., $5\\sqrt{13}$). Multi-step reasoning required.\n\n[statistics_probability] (MCQ + 1 SPR)\n- Topics: mean/median/mode, standard deviation (conceptual), probability (conditional from tables), data interpretation, scatter plots/line of best fit, sampling/bias\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Distractors: P(A|B) vs P(B|A) confusion, mean vs median, causal from correlational\n\n[advanced_math] (MCQ + 1 SPR)\n- Topics: quadratics (factoring, discriminant), polynomials (zeros, end behavior), exponentials (growth/decay), rational expressions, radical expressions, non-linear systems\n- Hardest questions on the test. Multi-step reasoning required. Elegant algebraic insight preferred.\n- All expressions must be factorable with integer coefficients. SPR answers are clean numbers.\n\nVERIFY EVERY ANSWER: Solve each problem step-by-step before finalizing correct_answer.\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "```json\n[\n {\n \"id\": \"sat11-math-m2-q13\",\n \"number\": 13,\n \"type\": \"geometry_trigonometry\",\n \"prompt\": \"In the xy-plane, a circle has a diameter with endpoints at $(2, -5)$ and $(10, 1)$. Which of the following is an equation of the circle?\",\n \"choices\": {\n \"A\": \"$(x - 6)^2 + (y + 2)^2 = 100$\",\n \"B\": \"$(x + 6)^2 + (y - 2)^2 = 25$\",\n \"C\": \"$(x - 4)^2 + (y + 2)^2 = 50$\",\n \"D\": \"$(x - 6)^2 + (y + 2)^2 = 25$\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"D\"\n },\n {\n \"id\": \"sat11-math-m2-q14\",\n \"number\": 14,\n \"type\": \"geometry_trigonometry\",\n \"prompt\": \"In right triangle $ABC$, the measure of angle $C$ is $90^\\\\circ$. If the length of side $AB$ is 13 and the length of side $BC$ is 5, what is the value of $\\\\tan(A)$?\",\n \"choices\": {\n \"A\": \"$\\\\frac{5}{13}$\",\n \"B\": \"$\\\\frac{12}{13}$\",\n \"C\": \"$\\\\frac{5}{12}$\",\n \"D\": \"$\\\\frac{12}{5}$\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m2-q15\",\n \"number\": 15,\n \"type\": \"geometry_trigonometry\",\n \"prompt\": \"A right circular cylinder and a right circular cone have the same radius and equal volumes. If the height of the cylinder is 12 inches, what is the height of the cone, in inches?\",\n \"choices\": {\n \"A\": \"36\",\n \"B\": \"12\",\n \"C\": \"4\",\n \"D\": \"24\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\"\n },\n {\n \"id\": \"sat11-math-m2-q16\",\n \"number\": 16,\n \"type\": \"geometry_trigonometry\",\n \"prompt\": \"In the xy-plane, $\\\\triangle PQR$ has vertices $P(0,0)$, $Q(6,0)$, and $R(3,4)$. If $\\\\triangle STU$ is similar to $\\\\triangle PQR$, with vertex $S$ at $(0,0)$ and vertex $T$ at $(9,0)$, what is the y-coordinate of vertex $U$?\",\n \"choices\": {\n \"A\": \"$\\\\frac{8}{3}$\",\n \"B\": \"6\",\n \"C\": \"4\",\n \"D\": \"7\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"B\"\n },\n {\n \"id\": \"sat11-math-m2-q17\",\n \"number\": 17,\n \"type\": \"geometry_trigonometry\",\n \"prompt\": \"A regular hexagon is inscribed in a circle. What is the measure in degrees of the acute angle formed between a side of the hexagon and a line tangent to the circle at a shared vertex?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": 30\n }\n]\n```" }, { "costUsd": 0, "model": "gemini-2.5-pro", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Math questions (of which 1 are student-produced-response and the rest are MCQ).\nType: statistics_probability | Test number: 11 | Module: 2 | First question number: 18\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-math-m2-qN\" where N is the sequential question number starting at 18.\nFor example, the first question in this batch has id \"sat11-math-m2-q18\".\n\nMCQ Schema: {\"id\":\"sat11-math-m2-qN\",\"number\":N,\"type\":\"statistics_probability\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\nSPR Schema: {\"id\":\"sat11-math-m2-qN\",\"number\":N,\"type\":\"statistics_probability\",\"prompt\":\"...\",\"answer_format\":\"student_produced\",\"correct_answer\":\"numeric_value\"}\nPlace SPR questions at the END of the array.\n\nFollow instructions for your type:\n\n[algebra] (MCQ + 1 SPR)\n- Topics: linear equations (1 var, 2 vars), systems of 2 linear equations, linear inequalities, absolute value\n- Mix word problems and abstract algebra. Use $...$ notation.\n- MCQ distractors: sign errors, arithmetic mistakes, setup errors, partial answers\n- SPR answers: clean integers or simple fractions\n\n[functions_analysis] (MCQ + 1 SPR)\n- Topics: function evaluation, domain/range, transformations (f(x)+k, f(x-h)), graphical analysis, composition f(g(x)), piecewise\n- At least one conceptual insight question (e.g., (p,q) and (q,p) on same graph)\n- Describe graphs in text. All computations yield clean results.\n\n[geometry_trigonometry] (MCQ + 1 SPR)\n- Topics: circle equations, Pythagorean theorem, special right triangles, coordinate geometry, trig ratios, volume/surface area, angle relationships\n- Optional \"visual\" field: {\"type\":\"coordinate_plane\",\"description\":\"...\"}\n- Radical answers OK (e.g., $5\\sqrt{13}$). Multi-step reasoning required.\n\n[statistics_probability] (MCQ + 1 SPR)\n- Topics: mean/median/mode, standard deviation (conceptual), probability (conditional from tables), data interpretation, scatter plots/line of best fit, sampling/bias\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Distractors: P(A|B) vs P(B|A) confusion, mean vs median, causal from correlational\n\n[advanced_math] (MCQ + 1 SPR)\n- Topics: quadratics (factoring, discriminant), polynomials (zeros, end behavior), exponentials (growth/decay), rational expressions, radical expressions, non-linear systems\n- Hardest questions on the test. Multi-step reasoning required. Elegant algebraic insight preferred.\n- All expressions must be factorable with integer coefficients. SPR answers are clean numbers.\n\nVERIFY EVERY ANSWER: Solve each problem step-by-step before finalizing correct_answer.\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "```json\n[\n {\n \"id\": \"sat11-math-m2-q18\",\n \"number\": 18,\n \"type\": \"statistics_probability\",\n \"prompt\": \"The scatterplot above shows the relationship between two variables, $x$ and $y$. A line of best fit for the data is also shown. For the data point $(10, 55)$, the actual $y$-value is 55. What is the $y$-value predicted by the line of best fit when $x=10$, and what is the value of the corresponding residual?\",\n \"visual\": {\n \"type\": \"coordinate_plane\",\n \"description\": \"A scatterplot with about 15 points that show a strong, positive, linear association. The line of best fit is drawn through the points. The line passes through approximately (0, 10) and (20, 90). The point (10, 55) is clearly plotted slightly above the line of best fit.\"\n },\n \"choices\": {\n \"A\": \"Predicted value: 50; Residual: -5\",\n \"B\": \"Predicted value: 55; Residual: 0\",\n \"C\": \"Predicted value: 50; Residual: 5\",\n \"D\": \"Predicted value: 5; Residual: 50\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m2-q19\",\n \"number\": 19,\n \"type\": \"statistics_probability\",\n \"prompt\": \"The table above summarizes the results of a survey of 200 smartphone users on their preferred mobile operating system and their primary use for the phone. Based on the table, what is the probability that a user who primarily uses their phone for communication prefers OS Y?\",\n \"visual\": {\n \"type\": \"table\",\n \"title\": \"Smartphone User Survey\",\n \"headers\": [\"Primary Use\", \"OS X\", \"OS Y\", \"Total\"],\n \"rows\": [\n [\"Communication\", 60, 40, 100],\n [\"Entertainment\", 70, 30, 100],\n [\"Total\", 130, 70, 200]\n ]\n },\n \"choices\": {\n \"A\": \"0.20\",\n \"B\": \"0.35\",\n \"C\": \"0.57\",\n \"D\": \"0.40\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"D\"\n },\n {\n \"id\": \"sat11-math-m2-q20\",\n \"number\": 20,\n \"type\": \"statistics_probability\",\n \"prompt\": \"A small business has 11 employees. The annual salaries, in thousands of dollars, are: 45, 45, 48, 50, 52, 55, 58, 60, 65, 70, and 300. Which of the following statements best compares the mean and median of these salaries?\",\n \"choices\": {\n \"A\": \"The mean is greater than the median.\",\n \"B\": \"The median is greater than the mean.\",\n \"C\": \"The mean and the median are equal.\",\n \"D\": \"The relationship between the mean and median cannot be determined.\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\"\n },\n {\n \"id\": \"sat11-math-m2-q21\",\n \"number\": 21,\n \"type\": \"statistics_probability\",\n \"prompt\": \"A market researcher for a national coffee chain wants to determine the average amount of money a customer spends per visit. Which of the following sampling methods is most likely to introduce a significant bias into the study?\",\n \"choices\": {\n \"A\": \"Selecting 100 customers at random from the company's national loyalty program database.\",\n \"B\": \"Surveying every customer who visits a single, specific store location between 8 a.m. and 9 a.m. on a Monday morning.\",\n \"C\": \"Dividing the chain's stores into regions and randomly selecting 5 stores from each region, then surveying all customers at those stores for one day.\",\n \"D\": \"Using a computer to randomly select 200 transaction times from the previous week and surveying the customer who made each of those transactions.\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"B\"\n },\n {\n \"id\": \"sat11-math-m2-q22\",\n \"number\": 22,\n \"type\": \"statistics_probability\",\n \"prompt\": \"At a conference of 300 attendees, 180 are from North America and the rest are from Europe. If 60% of the North American attendees are scientists and 40% of the European attendees are scientists, what is the probability that an attendee selected at random is a scientist? Express your answer as a decimal.\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": \"0.52\"\n }\n]\n```" }, { "costUsd": 0, "model": "gemini-2.5-pro", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Math questions (of which 2 are student-produced-response and the rest are MCQ).\nType: advanced_math | Test number: 11 | Module: 2 | First question number: 23\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-math-m2-qN\" where N is the sequential question number starting at 23.\nFor example, the first question in this batch has id \"sat11-math-m2-q23\".\n\nMCQ Schema: {\"id\":\"sat11-math-m2-qN\",\"number\":N,\"type\":\"advanced_math\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\nSPR Schema: {\"id\":\"sat11-math-m2-qN\",\"number\":N,\"type\":\"advanced_math\",\"prompt\":\"...\",\"answer_format\":\"student_produced\",\"correct_answer\":\"numeric_value\"}\nPlace SPR questions at the END of the array.\n\nFollow instructions for your type:\n\n[algebra] (MCQ + 2 SPR)\n- Topics: linear equations (1 var, 2 vars), systems of 2 linear equations, linear inequalities, absolute value\n- Mix word problems and abstract algebra. Use $...$ notation.\n- MCQ distractors: sign errors, arithmetic mistakes, setup errors, partial answers\n- SPR answers: clean integers or simple fractions\n\n[functions_analysis] (MCQ + 2 SPR)\n- Topics: function evaluation, domain/range, transformations (f(x)+k, f(x-h)), graphical analysis, composition f(g(x)), piecewise\n- At least one conceptual insight question (e.g., (p,q) and (q,p) on same graph)\n- Describe graphs in text. All computations yield clean results.\n\n[geometry_trigonometry] (MCQ + 2 SPR)\n- Topics: circle equations, Pythagorean theorem, special right triangles, coordinate geometry, trig ratios, volume/surface area, angle relationships\n- Optional \"visual\" field: {\"type\":\"coordinate_plane\",\"description\":\"...\"}\n- Radical answers OK (e.g., $5\\sqrt{13}$). Multi-step reasoning required.\n\n[statistics_probability] (MCQ + 2 SPR)\n- Topics: mean/median/mode, standard deviation (conceptual), probability (conditional from tables), data interpretation, scatter plots/line of best fit, sampling/bias\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Distractors: P(A|B) vs P(B|A) confusion, mean vs median, causal from correlational\n\n[advanced_math] (MCQ + 2 SPR)\n- Topics: quadratics (factoring, discriminant), polynomials (zeros, end behavior), exponentials (growth/decay), rational expressions, radical expressions, non-linear systems\n- Hardest questions on the test. Multi-step reasoning required. Elegant algebraic insight preferred.\n- All expressions must be factorable with integer coefficients. SPR answers are clean numbers.\n\nVERIFY EVERY ANSWER: Solve each problem step-by-step before finalizing correct_answer.\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "```json\n[\n {\n \"id\": \"sat11-math-m2-q23\",\n \"number\": 23,\n \"type\": \"advanced_math\",\n \"prompt\": \"The parabola with equation $y = x^2 - 5x + 11$ intersects the line with equation $y = 3x - 4$ at two points, $(x_1, y_1)$ and $(x_2, y_2)$. What is the value of $x_1 + x_2$?\",\n \"choices\": {\n \"A\": \"8\",\n \"B\": \"15\",\n \"C\": \"2\",\n \"D\": \"-2\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\"\n },\n {\n \"id\": \"sat11-math-m2-q24\",\n \"number\": 24,\n \"type\": \"advanced_math\",\n \"prompt\": \"The expression $\\\\frac{3x^2+5x-2}{x^2-4}$ can be written as $A + \\\\frac{B}{x-2}$, where $A$ and $B$ are constants and $x \\\\neq \\\\pm 2$. What is the value of $B$?\",\n \"choices\": {\n \"A\": \"3\",\n \"B\": \"5\",\n \"C\": \"8.5\",\n \"D\": \"4.5\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"D\"\n },\n {\n \"id\": \"sat11-math-m2-q25\",\n \"number\": 25,\n \"type\": \"advanced_math\",\n \"prompt\": \"For a function defined by $f(x) = c \\cdot p^x$, where $c$ and $p$ are positive constants, $f(1)=15$ and $f(3)=135$. What is the value of $p$?\",\n \"choices\": {\n \"A\": \"9\",\n \"B\": \"5\",\n \"C\": \"3\",\n \"D\": \"60\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m2-q26\",\n \"number\": 26,\n \"type\": \"advanced_math\",\n \"prompt\": \"If $x > 0$ and $2\\sqrt{x+2} = x+1$, what is the value of $x$?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": 2\n },\n {\n \"id\": \"sat11-math-m2-q27\",\n \"number\": 27,\n \"type\": \"advanced_math\",\n \"prompt\": \"In the $xy$-plane, the graphs of the functions $f(x) = -x^2+3$ and $g(x) = x^2+bx-1$ intersect at exactly one point. What is the value of $b$?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": 4\n }\n]\n```" } ], "results": [ { "attempts": 1, "index": 0, "output": { "costUsd": 0, "model": "gemini-2.5-pro", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 7 SAT Math questions (of which 2 are student-produced-response and the rest are MCQ).\nType: algebra | Test number: 11 | Module: 1 | First question number: 1\n\nMANDATORY: Your JSON array MUST contain exactly 7 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-math-m1-qN\" where N is the sequential question number starting at 1.\nFor example, the first question in this batch has id \"sat11-math-m1-q1\".\n\nMCQ Schema: {\"id\":\"sat11-math-m1-qN\",\"number\":N,\"type\":\"algebra\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\nSPR Schema: {\"id\":\"sat11-math-m1-qN\",\"number\":N,\"type\":\"algebra\",\"prompt\":\"...\",\"answer_format\":\"student_produced\",\"correct_answer\":\"numeric_value\"}\nPlace SPR questions at the END of the array.\n\nFollow instructions for your type:\n\n[algebra] (MCQ + 2 SPR)\n- Topics: linear equations (1 var, 2 vars), systems of 2 linear equations, linear inequalities, absolute value\n- Mix word problems and abstract algebra. Use $...$ notation.\n- MCQ distractors: sign errors, arithmetic mistakes, setup errors, partial answers\n- SPR answers: clean integers or simple fractions\n\n[functions_analysis] (MCQ + 2 SPR)\n- Topics: function evaluation, domain/range, transformations (f(x)+k, f(x-h)), graphical analysis, composition f(g(x)), piecewise\n- At least one conceptual insight question (e.g., (p,q) and (q,p) on same graph)\n- Describe graphs in text. All computations yield clean results.\n\n[geometry_trigonometry] (MCQ + 2 SPR)\n- Topics: circle equations, Pythagorean theorem, special right triangles, coordinate geometry, trig ratios, volume/surface area, angle relationships\n- Optional \"visual\" field: {\"type\":\"coordinate_plane\",\"description\":\"...\"}\n- Radical answers OK (e.g., $5\\sqrt{13}$). Multi-step reasoning required.\n\n[statistics_probability] (MCQ + 2 SPR)\n- Topics: mean/median/mode, standard deviation (conceptual), probability (conditional from tables), data interpretation, scatter plots/line of best fit, sampling/bias\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Distractors: P(A|B) vs P(B|A) confusion, mean vs median, causal from correlational\n\n[advanced_math] (MCQ + 2 SPR)\n- Topics: quadratics (factoring, discriminant), polynomials (zeros, end behavior), exponentials (growth/decay), rational expressions, radical expressions, non-linear systems\n- Hardest questions on the test. Multi-step reasoning required. Elegant algebraic insight preferred.\n- All expressions must be factorable with integer coefficients. SPR answers are clean numbers.\n\nVERIFY EVERY ANSWER: Solve each problem step-by-step before finalizing correct_answer.\nOutput ONLY a valid JSON array of EXACTLY 7 objects. No markdown. Count them before outputting.", "text": "```json\n[\n {\n \"id\": \"sat11-math-m1-q1\",\n \"number\": 1,\n \"type\": \"algebra\",\n \"prompt\": \"A taxi service charges a $2.50 pickup fee plus $1.75 per mile. If a passenger's total fare was $20.00, how many miles did the taxi travel?\",\n \"choices\": {\n \"A\": \"8\",\n \"B\": \"10\",\n \"C\": \"11.43\",\n \"D\": \"12.86\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"B\"\n },\n {\n \"id\": \"sat11-math-m1-q2\",\n \"number\": 2,\n \"type\": \"algebra\",\n \"prompt\": \"A small theater sells adult tickets for $12 each and child tickets for $8 each. If a total of 150 tickets were sold for a total of $1,560, how many adult tickets were sold?\",\n \"choices\": {\n \"A\": \"60\",\n \"B\": \"75\",\n \"C\": \"90\",\n \"D\": \"100\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m1-q3\",\n \"number\": 3,\n \"type\": \"algebra\",\n \"prompt\": \"Which of the following inequalities is equivalent to $7 - 2x > 15$?\",\n \"choices\": {\n \"A\": \"$x < -4$\",\n \"B\": \"$x > -4$\",\n \"C\": \"$x < 4$\",\n \"D\": \"$x > 4$\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\"\n },\n {\n \"id\": \"sat11-math-m1-q4\",\n \"number\": 4,\n \"type\": \"algebra\",\n \"prompt\": \"If $|\\frac{x}{2} - 3| = 4$, what is a possible value of $x$?\",\n \"choices\": {\n \"A\": \"2\",\n \"B\": \"7\",\n \"C\": \"-1\",\n \"D\": \"14\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"D\"\n },\n {\n \"id\": \"sat11-math-m1-q5\",\n \"number\": 5,\n \"type\": \"algebra\",\n \"prompt\": \"If $3x + 2y = 18$ and $2x + 3y = 17$, what is the value of $x+y$?\",\n \"choices\": {\n \"A\": \"1\",\n \"B\": \"4\",\n \"C\": \"7\",\n \"D\": \"35\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m1-q6\",\n \"number\": 6,\n \"type\": \"algebra\",\n \"prompt\": \"A rental company charges $40 for the first day and $25 for each additional day to rent a machine. If a rental bill for a machine was $215, for how many total days was it rented?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": \"8\"\n },\n {\n \"id\": \"sat11-math-m1-q7\",\n \"number\": 7,\n \"type\": \"algebra\",\n \"prompt\": \"The inequality $|n - 7| \\le 2$ describes the range of possible values for a variable $n$. If $n$ must be an integer, what is the largest possible value of $n$?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": \"9\"\n }\n]\n```" }, "status": "ok" }, { "attempts": 1, "index": 1, "output": { "costUsd": 0, "model": "gemini-2.5-pro", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Math questions (of which 1 are student-produced-response and the rest are MCQ).\nType: functions_analysis | Test number: 11 | Module: 1 | First question number: 8\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-math-m1-qN\" where N is the sequential question number starting at 8.\nFor example, the first question in this batch has id \"sat11-math-m1-q8\".\n\nMCQ Schema: {\"id\":\"sat11-math-m1-qN\",\"number\":N,\"type\":\"functions_analysis\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\nSPR Schema: {\"id\":\"sat11-math-m1-qN\",\"number\":N,\"type\":\"functions_analysis\",\"prompt\":\"...\",\"answer_format\":\"student_produced\",\"correct_answer\":\"numeric_value\"}\nPlace SPR questions at the END of the array.\n\nFollow instructions for your type:\n\n[algebra] (MCQ + 1 SPR)\n- Topics: linear equations (1 var, 2 vars), systems of 2 linear equations, linear inequalities, absolute value\n- Mix word problems and abstract algebra. Use $...$ notation.\n- MCQ distractors: sign errors, arithmetic mistakes, setup errors, partial answers\n- SPR answers: clean integers or simple fractions\n\n[functions_analysis] (MCQ + 1 SPR)\n- Topics: function evaluation, domain/range, transformations (f(x)+k, f(x-h)), graphical analysis, composition f(g(x)), piecewise\n- At least one conceptual insight question (e.g., (p,q) and (q,p) on same graph)\n- Describe graphs in text. All computations yield clean results.\n\n[geometry_trigonometry] (MCQ + 1 SPR)\n- Topics: circle equations, Pythagorean theorem, special right triangles, coordinate geometry, trig ratios, volume/surface area, angle relationships\n- Optional \"visual\" field: {\"type\":\"coordinate_plane\",\"description\":\"...\"}\n- Radical answers OK (e.g., $5\\sqrt{13}$). Multi-step reasoning required.\n\n[statistics_probability] (MCQ + 1 SPR)\n- Topics: mean/median/mode, standard deviation (conceptual), probability (conditional from tables), data interpretation, scatter plots/line of best fit, sampling/bias\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Distractors: P(A|B) vs P(B|A) confusion, mean vs median, causal from correlational\n\n[advanced_math] (MCQ + 1 SPR)\n- Topics: quadratics (factoring, discriminant), polynomials (zeros, end behavior), exponentials (growth/decay), rational expressions, radical expressions, non-linear systems\n- Hardest questions on the test. Multi-step reasoning required. Elegant algebraic insight preferred.\n- All expressions must be factorable with integer coefficients. SPR answers are clean numbers.\n\nVERIFY EVERY ANSWER: Solve each problem step-by-step before finalizing correct_answer.\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "```json\n[\n {\n \"id\": \"sat11-math-m1-q8\",\n \"number\": 8,\n \"type\": \"functions_analysis\",\n \"prompt\": \"For the function $f$ defined by $f(x) = 3x^2 - 5x + 1$, what is the value of $f(-2)$?\",\n \"choices\": {\n \"A\": \"3\",\n \"B\": \"-1\",\n \"C\": \"23\",\n \"D\": \"-21\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m1-q9\",\n \"number\": 9,\n \"type\": \"functions_analysis\",\n \"prompt\": \"The graph of the function $g$ in the $xy$-plane contains the point $(2, 5)$. Which of the following points must be on the graph of the function $h$ defined by $h(x) = g(x-3) + 4$?\",\n \"choices\": {\n \"A\": \"(-1, 9)\",\n \"B\": \"(5, 9)\",\n \"C\": \"(5, 1)\",\n \"D\": \"(-1, 1)\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"B\"\n },\n {\n \"id\": \"sat11-math-m1-q10\",\n \"number\": 10,\n \"type\": \"functions_analysis\",\n \"prompt\": \"Let the functions $f$ and $g$ be defined by $f(x) = x^2 + 4$ and $g(x) = \\\\sqrt{x-1}$. What is the domain of the composite function $h(x) = f(g(x))$?\",\n \"choices\": {\n \"A\": \"x \\\\ge 1\",\n \"B\": \"x \\\\ge -4\",\n \"C\": \"x \\\\le 1\",\n \"D\": \"All real numbers\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\"\n },\n {\n \"id\": \"sat11-math-m1-q11\",\n \"number\": 11,\n \"type\": \"functions_analysis\",\n \"prompt\": \"A function $f$ is defined as follows:\\n$f(x) = \\\\begin{cases} 2x + 7, & \\\\text{if } x < 3 \\\\\\\\ x^2 - 1, & \\\\text{if } x \\\\ge 3 \\\\end{cases}$\\nWhat is the value of $f(3)$?\",\n \"choices\": {\n \"A\": \"13\",\n \"B\": \"10\",\n \"C\": \"9\",\n \"D\": \"8\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"D\"\n },\n {\n \"id\": \"sat11-math-m1-q12\",\n \"number\": 12,\n \"type\": \"functions_analysis\",\n \"prompt\": \"The functions $g$ and $h$ are defined by $g(x) = 2x - 5$ and $h(x) = x^2 + 3$. What is the value of $g(h(2))$?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": \"9\"\n }\n]\n```" }, "status": "ok" }, { "attempts": 1, "index": 2, "output": { "costUsd": 0, "model": "gemini-2.5-pro", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Math questions (of which 1 are student-produced-response and the rest are MCQ).\nType: geometry_trigonometry | Test number: 11 | Module: 1 | First question number: 13\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-math-m1-qN\" where N is the sequential question number starting at 13.\nFor example, the first question in this batch has id \"sat11-math-m1-q13\".\n\nMCQ Schema: {\"id\":\"sat11-math-m1-qN\",\"number\":N,\"type\":\"geometry_trigonometry\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\nSPR Schema: {\"id\":\"sat11-math-m1-qN\",\"number\":N,\"type\":\"geometry_trigonometry\",\"prompt\":\"...\",\"answer_format\":\"student_produced\",\"correct_answer\":\"numeric_value\"}\nPlace SPR questions at the END of the array.\n\nFollow instructions for your type:\n\n[algebra] (MCQ + 1 SPR)\n- Topics: linear equations (1 var, 2 vars), systems of 2 linear equations, linear inequalities, absolute value\n- Mix word problems and abstract algebra. Use $...$ notation.\n- MCQ distractors: sign errors, arithmetic mistakes, setup errors, partial answers\n- SPR answers: clean integers or simple fractions\n\n[functions_analysis] (MCQ + 1 SPR)\n- Topics: function evaluation, domain/range, transformations (f(x)+k, f(x-h)), graphical analysis, composition f(g(x)), piecewise\n- At least one conceptual insight question (e.g., (p,q) and (q,p) on same graph)\n- Describe graphs in text. All computations yield clean results.\n\n[geometry_trigonometry] (MCQ + 1 SPR)\n- Topics: circle equations, Pythagorean theorem, special right triangles, coordinate geometry, trig ratios, volume/surface area, angle relationships\n- Optional \"visual\" field: {\"type\":\"coordinate_plane\",\"description\":\"...\"}\n- Radical answers OK (e.g., $5\\sqrt{13}$). Multi-step reasoning required.\n\n[statistics_probability] (MCQ + 1 SPR)\n- Topics: mean/median/mode, standard deviation (conceptual), probability (conditional from tables), data interpretation, scatter plots/line of best fit, sampling/bias\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Distractors: P(A|B) vs P(B|A) confusion, mean vs median, causal from correlational\n\n[advanced_math] (MCQ + 1 SPR)\n- Topics: quadratics (factoring, discriminant), polynomials (zeros, end behavior), exponentials (growth/decay), rational expressions, radical expressions, non-linear systems\n- Hardest questions on the test. Multi-step reasoning required. Elegant algebraic insight preferred.\n- All expressions must be factorable with integer coefficients. SPR answers are clean numbers.\n\nVERIFY EVERY ANSWER: Solve each problem step-by-step before finalizing correct_answer.\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "```json\n[\n {\n \"id\": \"sat11-math-m1-q13\",\n \"number\": 13,\n \"type\": \"geometry_trigonometry\",\n \"prompt\": \"In the xy-plane, a circle has a center at $(3, -5)$ and a radius of 7. Which of the following is the equation of the circle?\",\n \"choices\": {\n \"A\": \"$(x+3)^2 + (y-5)^2 = 7$\",\n \"B\": \"$(x-3)^2 + (y+5)^2 = 49$\",\n \"C\": \"$(x-3)^2 + (y+5)^2 = 7$\",\n \"D\": \"$(x+3)^2 + (y-5)^2 = 49$\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"B\"\n },\n {\n \"id\": \"sat11-math-m1-q14\",\n \"number\": 14,\n \"type\": \"geometry_trigonometry\",\n \"prompt\": \"In right triangle ABC, the right angle is at vertex C. If the length of side AC is 8 and the length of side BC is 15, what is the value of $\\\\cos(A)$?\",\n \"choices\": {\n \"A\": \"$8/17$\",\n \"B\": \"$15/17$\",\n \"C\": \"$8/15$\",\n \"D\": \"$17/8$\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\"\n },\n {\n \"id\": \"sat11-math-m1-q15\",\n \"number\": 15,\n \"type\": \"geometry_trigonometry\",\n \"prompt\": \"Two parallel lines are intersected by a transversal line. The measure of one of the resulting interior angles is $(3x + 10)^\\\\circ$ and the measure of its corresponding consecutive interior angle is $(5x + 30)^\\\\circ$. What is the value of $x$?\",\n \"choices\": {\n \"A\": \"10\",\n \"B\": \"17.5\",\n \"C\": \"20\",\n \"D\": \"40\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m1-q16\",\n \"number\": 16,\n \"type\": \"geometry_trigonometry\",\n \"prompt\": \"In right triangle $XYZ$, the measure of angle $Y$ is $90^\\\\circ$ and the measure of angle $X$ is $60^\\\\circ$. If side $XY$ (the side adjacent to angle $X$) has a length of 5, what is the length of the hypotenuse, $XZ$?\",\n \"choices\": {\n \"A\": \"$5\\\\sqrt{2}$\",\n \"B\": \"$5\\\\sqrt{3}$\",\n \"C\": \"$10\\\\sqrt{3}/3$\",\n \"D\": \"10\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"D\"\n },\n {\n \"id\": \"sat11-math-m1-q17\",\n \"number\": 17,\n \"type\": \"geometry_trigonometry\",\n \"prompt\": \"A rectangular prism has a height of 4 cm, a width of 3 cm, and a length of 12 cm. What is the length, in cm, of the space diagonal that connects two opposite vertices of the prism?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": \"13\"\n }\n]\n```" }, "status": "ok" }, { "attempts": 1, "index": 3, "output": { "costUsd": 0, "model": "gemini-2.5-pro", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Math questions (of which 1 are student-produced-response and the rest are MCQ).\nType: statistics_probability | Test number: 11 | Module: 1 | First question number: 18\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-math-m1-qN\" where N is the sequential question number starting at 18.\nFor example, the first question in this batch has id \"sat11-math-m1-q18\".\n\nMCQ Schema: {\"id\":\"sat11-math-m1-qN\",\"number\":N,\"type\":\"statistics_probability\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\nSPR Schema: {\"id\":\"sat11-math-m1-qN\",\"number\":N,\"type\":\"statistics_probability\",\"prompt\":\"...\",\"answer_format\":\"student_produced\",\"correct_answer\":\"numeric_value\"}\nPlace SPR questions at the END of the array.\n\nFollow instructions for your type:\n\n[algebra] (MCQ + 1 SPR)\n- Topics: linear equations (1 var, 2 vars), systems of 2 linear equations, linear inequalities, absolute value\n- Mix word problems and abstract algebra. Use $...$ notation.\n- MCQ distractors: sign errors, arithmetic mistakes, setup errors, partial answers\n- SPR answers: clean integers or simple fractions\n\n[functions_analysis] (MCQ + 1 SPR)\n- Topics: function evaluation, domain/range, transformations (f(x)+k, f(x-h)), graphical analysis, composition f(g(x)), piecewise\n- At least one conceptual insight question (e.g., (p,q) and (q,p) on same graph)\n- Describe graphs in text. All computations yield clean results.\n\n[geometry_trigonometry] (MCQ + 1 SPR)\n- Topics: circle equations, Pythagorean theorem, special right triangles, coordinate geometry, trig ratios, volume/surface area, angle relationships\n- Optional \"visual\" field: {\"type\":\"coordinate_plane\",\"description\":\"...\"}\n- Radical answers OK (e.g., $5\\sqrt{13}$). Multi-step reasoning required.\n\n[statistics_probability] (MCQ + 1 SPR)\n- Topics: mean/median/mode, standard deviation (conceptual), probability (conditional from tables), data interpretation, scatter plots/line of best fit, sampling/bias\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Distractors: P(A|B) vs P(B|A) confusion, mean vs median, causal from correlational\n\n[advanced_math] (MCQ + 1 SPR)\n- Topics: quadratics (factoring, discriminant), polynomials (zeros, end behavior), exponentials (growth/decay), rational expressions, radical expressions, non-linear systems\n- Hardest questions on the test. Multi-step reasoning required. Elegant algebraic insight preferred.\n- All expressions must be factorable with integer coefficients. SPR answers are clean numbers.\n\nVERIFY EVERY ANSWER: Solve each problem step-by-step before finalizing correct_answer.\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "```json\n[\n {\n \"id\": \"sat11-math-m1-q18\",\n \"number\": 18,\n \"type\": \"statistics_probability\",\n \"prompt\": \"A community garden manager models the daily water usage for a plot of tomato plants. The line of best fit for the model is $y = 0.25x + 1.5$, where $y$ is the amount of water used in liters and $x$ is the average daily temperature in degrees Celsius. What is the best interpretation of the slope of this line?\",\n \"choices\": {\n \"A\": \"The initial amount of water used, regardless of temperature, is 0.25 liters.\",\n \"B\": \"For every 1-degree Celsius increase in temperature, the plants require an additional 1.5 liters of water.\",\n \"C\": \"For every 1-degree Celsius increase in average daily temperature, the predicted water usage increases by 0.25 liters.\",\n \"D\": \"For every 0.25-degree Celsius increase in temperature, the water usage increases by 1 liter.\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m1-q19\",\n \"number\": 19,\n \"type\": \"statistics_probability\",\n \"visual\": {\n \"type\": \"table\",\n \"title\": \"Instrument Played by Grade Level\",\n \"headers\": [\"\", \"Strings\", \"Woodwind\", \"Total\"],\n \"rows\": [\n [\"Grade 9\", 18, 22, 40],\n [\"Grade 10\", 24, 16, 40],\n [\"Total\", 42, 38, 80]\n ]\n },\n \"prompt\": \"The table above shows the distribution of 80 students in a music program by grade level and the type of instrument they play. If a student who plays a strings instrument is selected at random, what is the probability that the student is in Grade 9?\",\n \"choices\": {\n \"A\": \"$\\\\frac{18}{80}$\",\n \"B\": \"$\\\\frac{18}{40}$\",\n \"C\": \"$\\\\frac{40}{80}$\",\n \"D\": \"$\\\\frac{18}{42}$\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"D\"\n },\n {\n \"id\": \"sat11-math-m1-q20\",\n \"number\": 20,\n \"type\": \"statistics_probability\",\n \"prompt\": \"The list shows the number of points scored by a basketball player in each of the last 7 games: 15, 21, 18, 24, 19, 35, 17. Which statement accurately compares the mean and median of these scores?\",\n \"choices\": {\n \"A\": \"The mean is greater than the median.\",\n \"B\": \"The median is greater than the mean.\",\n \"C\": \"The mean and the median are equal.\",\n \"D\": \"The relationship cannot be determined from the information given.\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\"\n },\n {\n \"id\": \"sat11-math-m1-q21\",\n \"number\": 21,\n \"type\": \"statistics_probability\",\n \"visual\": {\n \"type\": \"table\",\n \"title\": \"City Employee Departments\",\n \"headers\": [\"Department\", \"Number of Employees\"],\n \"rows\": [\n [\"Public Works\", 450],\n [\"Parks & Recreation\", 200],\n [\"Administration\", 150],\n [\"Safety\", 700]\n ]\n },\n \"prompt\": \"A researcher wants to survey employees about workplace satisfaction in the city described in the table. Which sampling method is most likely to introduce bias?\",\n \"choices\": {\n \"A\": \"Using a random number generator to select 100 employees from a full alphabetical list of all city employees.\",\n \"B\": \"Leaving survey forms at the entrance of the Safety department building and using the first 100 that are returned.\",\n \"C\": \"Randomly selecting 50 employees from the Public Works department.\",\n \"D\": \"Creating a stratified sample by selecting employees from each department in proportion to the department's size.\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"B\"\n },\n {\n \"id\": \"sat11-math-m1-q22\",\n \"number\": 22,\n \"type\": \"statistics_probability\",\n \"prompt\": \"A dataset of 14 numbers has a mean of 68. A fifteenth number is added to the dataset, and the new mean is 70. What is the value of the number that was added to the dataset?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": 98\n }\n]\n```" }, "status": "ok" }, { "attempts": 1, "index": 4, "output": { "costUsd": 0, "model": "gemini-2.5-pro", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Math questions (of which 2 are student-produced-response and the rest are MCQ).\nType: advanced_math | Test number: 11 | Module: 1 | First question number: 23\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-math-m1-qN\" where N is the sequential question number starting at 23.\nFor example, the first question in this batch has id \"sat11-math-m1-q23\".\n\nMCQ Schema: {\"id\":\"sat11-math-m1-qN\",\"number\":N,\"type\":\"advanced_math\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\nSPR Schema: {\"id\":\"sat11-math-m1-qN\",\"number\":N,\"type\":\"advanced_math\",\"prompt\":\"...\",\"answer_format\":\"student_produced\",\"correct_answer\":\"numeric_value\"}\nPlace SPR questions at the END of the array.\n\nFollow instructions for your type:\n\n[algebra] (MCQ + 2 SPR)\n- Topics: linear equations (1 var, 2 vars), systems of 2 linear equations, linear inequalities, absolute value\n- Mix word problems and abstract algebra. Use $...$ notation.\n- MCQ distractors: sign errors, arithmetic mistakes, setup errors, partial answers\n- SPR answers: clean integers or simple fractions\n\n[functions_analysis] (MCQ + 2 SPR)\n- Topics: function evaluation, domain/range, transformations (f(x)+k, f(x-h)), graphical analysis, composition f(g(x)), piecewise\n- At least one conceptual insight question (e.g., (p,q) and (q,p) on same graph)\n- Describe graphs in text. All computations yield clean results.\n\n[geometry_trigonometry] (MCQ + 2 SPR)\n- Topics: circle equations, Pythagorean theorem, special right triangles, coordinate geometry, trig ratios, volume/surface area, angle relationships\n- Optional \"visual\" field: {\"type\":\"coordinate_plane\",\"description\":\"...\"}\n- Radical answers OK (e.g., $5\\sqrt{13}$). Multi-step reasoning required.\n\n[statistics_probability] (MCQ + 2 SPR)\n- Topics: mean/median/mode, standard deviation (conceptual), probability (conditional from tables), data interpretation, scatter plots/line of best fit, sampling/bias\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Distractors: P(A|B) vs P(B|A) confusion, mean vs median, causal from correlational\n\n[advanced_math] (MCQ + 2 SPR)\n- Topics: quadratics (factoring, discriminant), polynomials (zeros, end behavior), exponentials (growth/decay), rational expressions, radical expressions, non-linear systems\n- Hardest questions on the test. Multi-step reasoning required. Elegant algebraic insight preferred.\n- All expressions must be factorable with integer coefficients. SPR answers are clean numbers.\n\nVERIFY EVERY ANSWER: Solve each problem step-by-step before finalizing correct_answer.\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "```json\n[\n {\n \"id\": \"sat11-math-m1-q23\",\n \"number\": 23,\n \"type\": \"advanced_math\",\n \"prompt\": \"In the xy-plane, a parabola with the equation $y = x^2 + bx + c$ and a line with the equation $y = -x + k$ intersect at exactly two points, $(2, 5)$ and $(5, 2)$. What is the value of $b$?\",\n \"choices\": {\n \"A\": \"-8\",\n \"B\": \"-6\",\n \"C\": \"2\",\n \"D\": \"7\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"B\"\n },\n {\n \"id\": \"sat11-math-m1-q24\",\n \"number\": 24,\n \"type\": \"advanced_math\",\n \"prompt\": \"Which of the following expressions is equivalent to $\\\\frac{2x^2-11x+10}{x-4}$ for $x>4$?\",\n \"choices\": {\n \"A\": \"$2x-3 - \\\\frac{2}{x-4}$\",\n \"B\": \"$2x+3 + \\\\frac{22}{x-4}$\",\n \"C\": \"$2x-19 + \\\\frac{86}{x-4}$\",\n \"D\": \"$2x-3 + \\\\frac{22}{x-4}$\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\"\n },\n {\n \"id\": \"sat11-math-m1-q25\",\n \"number\": 25,\n \"type\": \"advanced_math\",\n \"prompt\": \"The function $f$ is defined by $f(x) = k(x+a)(x+b)$, where $k$, $a$, and $b$ are constants. The graph of $y=f(x)$ in the xy-plane is a parabola with x-intercepts at $-3$ and $5$. The vertex of the parabola is $(1, -32)$. What is the value of $k$?\",\n \"choices\": {\n \"A\": \"-2\",\n \"B\": \"1\",\n \"C\": \"2\",\n \"D\": \"8\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m1-q26\",\n \"number\": 26,\n \"type\": \"advanced_math\",\n \"prompt\": \"If $x > 0$ and $\\\\sqrt{3\\\\sqrt{x}} = \\\\sqrt[n]{3x}$ for all values of $x$, what is the value of $n$?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": 4\n },\n {\n \"id\": \"sat11-math-m1-q27\",\n \"number\": 27,\n \"type\": \"advanced_math\",\n \"prompt\": \"For the quadratic equation $kx^2 - 8x + 2 = 0$, the sum of the solutions is twice the product of the solutions. What is the value of $k$?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": 2\n }\n]\n```" }, "status": "ok" }, { "attempts": 1, "index": 5, "output": { "costUsd": 0, "model": "gemini-2.5-pro", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 7 SAT Math questions (of which 2 are student-produced-response and the rest are MCQ).\nType: algebra | Test number: 11 | Module: 2 | First question number: 1\n\nMANDATORY: Your JSON array MUST contain exactly 7 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-math-m2-qN\" where N is the sequential question number starting at 1.\nFor example, the first question in this batch has id \"sat11-math-m2-q1\".\n\nMCQ Schema: {\"id\":\"sat11-math-m2-qN\",\"number\":N,\"type\":\"algebra\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\nSPR Schema: {\"id\":\"sat11-math-m2-qN\",\"number\":N,\"type\":\"algebra\",\"prompt\":\"...\",\"answer_format\":\"student_produced\",\"correct_answer\":\"numeric_value\"}\nPlace SPR questions at the END of the array.\n\nFollow instructions for your type:\n\n[algebra] (MCQ + 2 SPR)\n- Topics: linear equations (1 var, 2 vars), systems of 2 linear equations, linear inequalities, absolute value\n- Mix word problems and abstract algebra. Use $...$ notation.\n- MCQ distractors: sign errors, arithmetic mistakes, setup errors, partial answers\n- SPR answers: clean integers or simple fractions\n\n[functions_analysis] (MCQ + 2 SPR)\n- Topics: function evaluation, domain/range, transformations (f(x)+k, f(x-h)), graphical analysis, composition f(g(x)), piecewise\n- At least one conceptual insight question (e.g., (p,q) and (q,p) on same graph)\n- Describe graphs in text. All computations yield clean results.\n\n[geometry_trigonometry] (MCQ + 2 SPR)\n- Topics: circle equations, Pythagorean theorem, special right triangles, coordinate geometry, trig ratios, volume/surface area, angle relationships\n- Optional \"visual\" field: {\"type\":\"coordinate_plane\",\"description\":\"...\"}\n- Radical answers OK (e.g., $5\\sqrt{13}$). Multi-step reasoning required.\n\n[statistics_probability] (MCQ + 2 SPR)\n- Topics: mean/median/mode, standard deviation (conceptual), probability (conditional from tables), data interpretation, scatter plots/line of best fit, sampling/bias\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Distractors: P(A|B) vs P(B|A) confusion, mean vs median, causal from correlational\n\n[advanced_math] (MCQ + 2 SPR)\n- Topics: quadratics (factoring, discriminant), polynomials (zeros, end behavior), exponentials (growth/decay), rational expressions, radical expressions, non-linear systems\n- Hardest questions on the test. Multi-step reasoning required. Elegant algebraic insight preferred.\n- All expressions must be factorable with integer coefficients. SPR answers are clean numbers.\n\nVERIFY EVERY ANSWER: Solve each problem step-by-step before finalizing correct_answer.\nOutput ONLY a valid JSON array of EXACTLY 7 objects. No markdown. Count them before outputting.", "text": "```json\n[\n {\n \"id\": \"sat11-math-m2-q1\",\n \"number\": 1,\n \"type\": \"algebra\",\n \"prompt\": \"If $\\\\frac{1}{3}(6x + 15) - 4 = 9$, what is the value of $x$?\",\n \"choices\": {\n \"A\": \"-1\",\n \"B\": \"4\",\n \"C\": \"5\",\n \"D\": \"8\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"B\"\n },\n {\n \"id\": \"sat11-math-m2-q2\",\n \"number\": 2,\n \"type\": \"algebra\",\n \"prompt\": \"A gym offers two membership plans. Plan A has a one-time initiation fee of $50 and a monthly fee of $20. Plan B has no initiation fee and a monthly fee of $30. After how many months will the total cost of both plans be equal?\",\n \"choices\": {\n \"A\": \"1.5\",\n \"B\": \"2.5\",\n \"C\": \"5\",\n \"D\": \"10\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m2-q3\",\n \"number\": 3,\n \"type\": \"algebra\",\n \"prompt\": \"Which of the following is a possible value of $y$ if $-4(y - 2) > 12$?\",\n \"choices\": {\n \"A\": \"-2\",\n \"B\": \"-1\",\n \"C\": \"0\",\n \"D\": \"2\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\"\n },\n {\n \"id\": \"sat11-math-m2-q4\",\n \"number\": 4,\n \"type\": \"algebra\",\n \"prompt\": \"If $|2k - 3| = 7$, what is the sum of the two possible values for $k$?\",\n \"choices\": {\n \"A\": \"-2\",\n \"B\": \"0\",\n \"C\": \"3\",\n \"D\": \"5\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m2-q5\",\n \"number\": 5,\n \"type\": \"algebra\",\n \"prompt\": \"If $5x - 2y = 11$ and $x + 3y = 9$, what is the value of $x$?\",\n \"choices\": {\n \"A\": \"-1\",\n \"B\": \"2\",\n \"C\": \"2.5\",\n \"D\": \"3\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"D\"\n },\n {\n \"id\": \"sat11-math-m2-q6\",\n \"number\": 6,\n \"type\": \"algebra\",\n \"prompt\": \"A rental car company charges a flat daily fee plus a fee per mile driven. A customer who drove 100 miles was charged $80. Another customer drove 150 miles on the same day and was charged $95. What is the company's flat daily fee in dollars?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": 50\n },\n {\n \"id\": \"sat11-math-m2-q7\",\n \"number\": 7,\n \"type\": \"algebra\",\n \"prompt\": \"The inequality $|x - 5| \\\\leq 2$ defines a range of values for $x$. If $x$ must be an integer, what is the sum of all possible values of $x$?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": 25\n }\n]\n```" }, "status": "ok" }, { "attempts": 1, "index": 6, "output": { "costUsd": 0, "model": "gemini-2.5-pro", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Math questions (of which 1 are student-produced-response and the rest are MCQ).\nType: functions_analysis | Test number: 11 | Module: 2 | First question number: 8\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-math-m2-qN\" where N is the sequential question number starting at 8.\nFor example, the first question in this batch has id \"sat11-math-m2-q8\".\n\nMCQ Schema: {\"id\":\"sat11-math-m2-qN\",\"number\":N,\"type\":\"functions_analysis\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\nSPR Schema: {\"id\":\"sat11-math-m2-qN\",\"number\":N,\"type\":\"functions_analysis\",\"prompt\":\"...\",\"answer_format\":\"student_produced\",\"correct_answer\":\"numeric_value\"}\nPlace SPR questions at the END of the array.\n\nFollow instructions for your type:\n\n[algebra] (MCQ + 1 SPR)\n- Topics: linear equations (1 var, 2 vars), systems of 2 linear equations, linear inequalities, absolute value\n- Mix word problems and abstract algebra. Use $...$ notation.\n- MCQ distractors: sign errors, arithmetic mistakes, setup errors, partial answers\n- SPR answers: clean integers or simple fractions\n\n[functions_analysis] (MCQ + 1 SPR)\n- Topics: function evaluation, domain/range, transformations (f(x)+k, f(x-h)), graphical analysis, composition f(g(x)), piecewise\n- At least one conceptual insight question (e.g., (p,q) and (q,p) on same graph)\n- Describe graphs in text. All computations yield clean results.\n\n[geometry_trigonometry] (MCQ + 1 SPR)\n- Topics: circle equations, Pythagorean theorem, special right triangles, coordinate geometry, trig ratios, volume/surface area, angle relationships\n- Optional \"visual\" field: {\"type\":\"coordinate_plane\",\"description\":\"...\"}\n- Radical answers OK (e.g., $5\\sqrt{13}$). Multi-step reasoning required.\n\n[statistics_probability] (MCQ + 1 SPR)\n- Topics: mean/median/mode, standard deviation (conceptual), probability (conditional from tables), data interpretation, scatter plots/line of best fit, sampling/bias\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Distractors: P(A|B) vs P(B|A) confusion, mean vs median, causal from correlational\n\n[advanced_math] (MCQ + 1 SPR)\n- Topics: quadratics (factoring, discriminant), polynomials (zeros, end behavior), exponentials (growth/decay), rational expressions, radical expressions, non-linear systems\n- Hardest questions on the test. Multi-step reasoning required. Elegant algebraic insight preferred.\n- All expressions must be factorable with integer coefficients. SPR answers are clean numbers.\n\nVERIFY EVERY ANSWER: Solve each problem step-by-step before finalizing correct_answer.\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "```json\n[\n {\n \"id\": \"sat11-math-m2-q8\",\n \"number\": 8,\n \"type\": \"functions_analysis\",\n \"prompt\": \"Let the function $h$ be defined by $h(x) = x^2 - 3x$. If $h(a+2) = 4$ for some constant $a$, what is one possible value of $a$?\",\n \"choices\": {\n \"A\": \"2\",\n \"B\": \"-1\",\n \"C\": \"0\",\n \"D\": \"4\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\"\n },\n {\n \"id\": \"sat11-math-m2-q9\",\n \"number\": 9,\n \"type\": \"functions_analysis\",\n \"prompt\": \"The graph of a function $f$ in the $xy$-plane passes through the point $(0, 3)$. A second function, $g$, is defined by the transformation $g(x) = f(x-2) + 1$. Which of the following points must be on the graph of $g$?\",\n \"choices\": {\n \"A\": \"(-2, 4)\",\n \"B\": \"(2, 4)\",\n \"C\": \"(2, 2)\",\n \"D\": \"(-2, 2)\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"B\"\n },\n {\n \"id\": \"sat11-math-m2-q10\",\n \"number\": 10,\n \"type\": \"functions_analysis\",\n \"prompt\": \"A function $f$ is defined as $f(x) = \\\\begin{cases} 2x + 7 & \\\\text{if } x \\\\le 1 \\\\\\\\ x^2 - 4 & \\\\text{if } x > 1 \\\\end{cases}$. What is the value of $f(-3) + f(3)$?\",\n \"choices\": {\n \"A\": \"14\",\n \"B\": \"10\",\n \"C\": \"6\",\n \"D\": \"4\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m2-q11\",\n \"number\": 11,\n \"type\": \"functions_analysis\",\n \"prompt\": \"The function $g$ is defined by $g(x) = \\\\frac{x+3}{\\\\sqrt{4-x}}$. For the function's output to be a real number, which of the following inequalities describes all possible values of the input $x$?\",\n \"choices\": {\n \"A\": \"$x > 4$\",\n \"B\": \"$x \\\\ge 4$\",\n \"C\": \"$x \\\\le 4$\",\n \"D\": \"$x < 4$\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"D\"\n },\n {\n \"id\": \"sat11-math-m2-q12\",\n \"number\": 12,\n \"type\": \"functions_analysis\",\n \"prompt\": \"Let the functions $f$ and $g$ be defined by $f(x) = 2x - 5$ and $g(x) = x^2 + k$, where $k$ is a constant. If $f(g(1)) = -3$, what is the value of $k$?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": \"0\"\n }\n]\n```" }, "status": "ok" }, { "attempts": 1, "index": 7, "output": { "costUsd": 0, "model": "gemini-2.5-pro", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Math questions (of which 1 are student-produced-response and the rest are MCQ).\nType: geometry_trigonometry | Test number: 11 | Module: 2 | First question number: 13\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-math-m2-qN\" where N is the sequential question number starting at 13.\nFor example, the first question in this batch has id \"sat11-math-m2-q13\".\n\nMCQ Schema: {\"id\":\"sat11-math-m2-qN\",\"number\":N,\"type\":\"geometry_trigonometry\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\nSPR Schema: {\"id\":\"sat11-math-m2-qN\",\"number\":N,\"type\":\"geometry_trigonometry\",\"prompt\":\"...\",\"answer_format\":\"student_produced\",\"correct_answer\":\"numeric_value\"}\nPlace SPR questions at the END of the array.\n\nFollow instructions for your type:\n\n[algebra] (MCQ + 1 SPR)\n- Topics: linear equations (1 var, 2 vars), systems of 2 linear equations, linear inequalities, absolute value\n- Mix word problems and abstract algebra. Use $...$ notation.\n- MCQ distractors: sign errors, arithmetic mistakes, setup errors, partial answers\n- SPR answers: clean integers or simple fractions\n\n[functions_analysis] (MCQ + 1 SPR)\n- Topics: function evaluation, domain/range, transformations (f(x)+k, f(x-h)), graphical analysis, composition f(g(x)), piecewise\n- At least one conceptual insight question (e.g., (p,q) and (q,p) on same graph)\n- Describe graphs in text. All computations yield clean results.\n\n[geometry_trigonometry] (MCQ + 1 SPR)\n- Topics: circle equations, Pythagorean theorem, special right triangles, coordinate geometry, trig ratios, volume/surface area, angle relationships\n- Optional \"visual\" field: {\"type\":\"coordinate_plane\",\"description\":\"...\"}\n- Radical answers OK (e.g., $5\\sqrt{13}$). Multi-step reasoning required.\n\n[statistics_probability] (MCQ + 1 SPR)\n- Topics: mean/median/mode, standard deviation (conceptual), probability (conditional from tables), data interpretation, scatter plots/line of best fit, sampling/bias\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Distractors: P(A|B) vs P(B|A) confusion, mean vs median, causal from correlational\n\n[advanced_math] (MCQ + 1 SPR)\n- Topics: quadratics (factoring, discriminant), polynomials (zeros, end behavior), exponentials (growth/decay), rational expressions, radical expressions, non-linear systems\n- Hardest questions on the test. Multi-step reasoning required. Elegant algebraic insight preferred.\n- All expressions must be factorable with integer coefficients. SPR answers are clean numbers.\n\nVERIFY EVERY ANSWER: Solve each problem step-by-step before finalizing correct_answer.\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "```json\n[\n {\n \"id\": \"sat11-math-m2-q13\",\n \"number\": 13,\n \"type\": \"geometry_trigonometry\",\n \"prompt\": \"In the xy-plane, a circle has a diameter with endpoints at $(2, -5)$ and $(10, 1)$. Which of the following is an equation of the circle?\",\n \"choices\": {\n \"A\": \"$(x - 6)^2 + (y + 2)^2 = 100$\",\n \"B\": \"$(x + 6)^2 + (y - 2)^2 = 25$\",\n \"C\": \"$(x - 4)^2 + (y + 2)^2 = 50$\",\n \"D\": \"$(x - 6)^2 + (y + 2)^2 = 25$\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"D\"\n },\n {\n \"id\": \"sat11-math-m2-q14\",\n \"number\": 14,\n \"type\": \"geometry_trigonometry\",\n \"prompt\": \"In right triangle $ABC$, the measure of angle $C$ is $90^\\\\circ$. If the length of side $AB$ is 13 and the length of side $BC$ is 5, what is the value of $\\\\tan(A)$?\",\n \"choices\": {\n \"A\": \"$\\\\frac{5}{13}$\",\n \"B\": \"$\\\\frac{12}{13}$\",\n \"C\": \"$\\\\frac{5}{12}$\",\n \"D\": \"$\\\\frac{12}{5}$\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m2-q15\",\n \"number\": 15,\n \"type\": \"geometry_trigonometry\",\n \"prompt\": \"A right circular cylinder and a right circular cone have the same radius and equal volumes. If the height of the cylinder is 12 inches, what is the height of the cone, in inches?\",\n \"choices\": {\n \"A\": \"36\",\n \"B\": \"12\",\n \"C\": \"4\",\n \"D\": \"24\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\"\n },\n {\n \"id\": \"sat11-math-m2-q16\",\n \"number\": 16,\n \"type\": \"geometry_trigonometry\",\n \"prompt\": \"In the xy-plane, $\\\\triangle PQR$ has vertices $P(0,0)$, $Q(6,0)$, and $R(3,4)$. If $\\\\triangle STU$ is similar to $\\\\triangle PQR$, with vertex $S$ at $(0,0)$ and vertex $T$ at $(9,0)$, what is the y-coordinate of vertex $U$?\",\n \"choices\": {\n \"A\": \"$\\\\frac{8}{3}$\",\n \"B\": \"6\",\n \"C\": \"4\",\n \"D\": \"7\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"B\"\n },\n {\n \"id\": \"sat11-math-m2-q17\",\n \"number\": 17,\n \"type\": \"geometry_trigonometry\",\n \"prompt\": \"A regular hexagon is inscribed in a circle. What is the measure in degrees of the acute angle formed between a side of the hexagon and a line tangent to the circle at a shared vertex?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": 30\n }\n]\n```" }, "status": "ok" }, { "attempts": 1, "index": 8, "output": { "costUsd": 0, "model": "gemini-2.5-pro", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Math questions (of which 1 are student-produced-response and the rest are MCQ).\nType: statistics_probability | Test number: 11 | Module: 2 | First question number: 18\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-math-m2-qN\" where N is the sequential question number starting at 18.\nFor example, the first question in this batch has id \"sat11-math-m2-q18\".\n\nMCQ Schema: {\"id\":\"sat11-math-m2-qN\",\"number\":N,\"type\":\"statistics_probability\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\nSPR Schema: {\"id\":\"sat11-math-m2-qN\",\"number\":N,\"type\":\"statistics_probability\",\"prompt\":\"...\",\"answer_format\":\"student_produced\",\"correct_answer\":\"numeric_value\"}\nPlace SPR questions at the END of the array.\n\nFollow instructions for your type:\n\n[algebra] (MCQ + 1 SPR)\n- Topics: linear equations (1 var, 2 vars), systems of 2 linear equations, linear inequalities, absolute value\n- Mix word problems and abstract algebra. Use $...$ notation.\n- MCQ distractors: sign errors, arithmetic mistakes, setup errors, partial answers\n- SPR answers: clean integers or simple fractions\n\n[functions_analysis] (MCQ + 1 SPR)\n- Topics: function evaluation, domain/range, transformations (f(x)+k, f(x-h)), graphical analysis, composition f(g(x)), piecewise\n- At least one conceptual insight question (e.g., (p,q) and (q,p) on same graph)\n- Describe graphs in text. All computations yield clean results.\n\n[geometry_trigonometry] (MCQ + 1 SPR)\n- Topics: circle equations, Pythagorean theorem, special right triangles, coordinate geometry, trig ratios, volume/surface area, angle relationships\n- Optional \"visual\" field: {\"type\":\"coordinate_plane\",\"description\":\"...\"}\n- Radical answers OK (e.g., $5\\sqrt{13}$). Multi-step reasoning required.\n\n[statistics_probability] (MCQ + 1 SPR)\n- Topics: mean/median/mode, standard deviation (conceptual), probability (conditional from tables), data interpretation, scatter plots/line of best fit, sampling/bias\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Distractors: P(A|B) vs P(B|A) confusion, mean vs median, causal from correlational\n\n[advanced_math] (MCQ + 1 SPR)\n- Topics: quadratics (factoring, discriminant), polynomials (zeros, end behavior), exponentials (growth/decay), rational expressions, radical expressions, non-linear systems\n- Hardest questions on the test. Multi-step reasoning required. Elegant algebraic insight preferred.\n- All expressions must be factorable with integer coefficients. SPR answers are clean numbers.\n\nVERIFY EVERY ANSWER: Solve each problem step-by-step before finalizing correct_answer.\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "```json\n[\n {\n \"id\": \"sat11-math-m2-q18\",\n \"number\": 18,\n \"type\": \"statistics_probability\",\n \"prompt\": \"The scatterplot above shows the relationship between two variables, $x$ and $y$. A line of best fit for the data is also shown. For the data point $(10, 55)$, the actual $y$-value is 55. What is the $y$-value predicted by the line of best fit when $x=10$, and what is the value of the corresponding residual?\",\n \"visual\": {\n \"type\": \"coordinate_plane\",\n \"description\": \"A scatterplot with about 15 points that show a strong, positive, linear association. The line of best fit is drawn through the points. The line passes through approximately (0, 10) and (20, 90). The point (10, 55) is clearly plotted slightly above the line of best fit.\"\n },\n \"choices\": {\n \"A\": \"Predicted value: 50; Residual: -5\",\n \"B\": \"Predicted value: 55; Residual: 0\",\n \"C\": \"Predicted value: 50; Residual: 5\",\n \"D\": \"Predicted value: 5; Residual: 50\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m2-q19\",\n \"number\": 19,\n \"type\": \"statistics_probability\",\n \"prompt\": \"The table above summarizes the results of a survey of 200 smartphone users on their preferred mobile operating system and their primary use for the phone. Based on the table, what is the probability that a user who primarily uses their phone for communication prefers OS Y?\",\n \"visual\": {\n \"type\": \"table\",\n \"title\": \"Smartphone User Survey\",\n \"headers\": [\"Primary Use\", \"OS X\", \"OS Y\", \"Total\"],\n \"rows\": [\n [\"Communication\", 60, 40, 100],\n [\"Entertainment\", 70, 30, 100],\n [\"Total\", 130, 70, 200]\n ]\n },\n \"choices\": {\n \"A\": \"0.20\",\n \"B\": \"0.35\",\n \"C\": \"0.57\",\n \"D\": \"0.40\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"D\"\n },\n {\n \"id\": \"sat11-math-m2-q20\",\n \"number\": 20,\n \"type\": \"statistics_probability\",\n \"prompt\": \"A small business has 11 employees. The annual salaries, in thousands of dollars, are: 45, 45, 48, 50, 52, 55, 58, 60, 65, 70, and 300. Which of the following statements best compares the mean and median of these salaries?\",\n \"choices\": {\n \"A\": \"The mean is greater than the median.\",\n \"B\": \"The median is greater than the mean.\",\n \"C\": \"The mean and the median are equal.\",\n \"D\": \"The relationship between the mean and median cannot be determined.\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\"\n },\n {\n \"id\": \"sat11-math-m2-q21\",\n \"number\": 21,\n \"type\": \"statistics_probability\",\n \"prompt\": \"A market researcher for a national coffee chain wants to determine the average amount of money a customer spends per visit. Which of the following sampling methods is most likely to introduce a significant bias into the study?\",\n \"choices\": {\n \"A\": \"Selecting 100 customers at random from the company's national loyalty program database.\",\n \"B\": \"Surveying every customer who visits a single, specific store location between 8 a.m. and 9 a.m. on a Monday morning.\",\n \"C\": \"Dividing the chain's stores into regions and randomly selecting 5 stores from each region, then surveying all customers at those stores for one day.\",\n \"D\": \"Using a computer to randomly select 200 transaction times from the previous week and surveying the customer who made each of those transactions.\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"B\"\n },\n {\n \"id\": \"sat11-math-m2-q22\",\n \"number\": 22,\n \"type\": \"statistics_probability\",\n \"prompt\": \"At a conference of 300 attendees, 180 are from North America and the rest are from Europe. If 60% of the North American attendees are scientists and 40% of the European attendees are scientists, what is the probability that an attendee selected at random is a scientist? Express your answer as a decimal.\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": \"0.52\"\n }\n]\n```" }, "status": "ok" }, { "attempts": 1, "index": 9, "output": { "costUsd": 0, "model": "gemini-2.5-pro", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Math questions (of which 2 are student-produced-response and the rest are MCQ).\nType: advanced_math | Test number: 11 | Module: 2 | First question number: 23\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-math-m2-qN\" where N is the sequential question number starting at 23.\nFor example, the first question in this batch has id \"sat11-math-m2-q23\".\n\nMCQ Schema: {\"id\":\"sat11-math-m2-qN\",\"number\":N,\"type\":\"advanced_math\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\nSPR Schema: {\"id\":\"sat11-math-m2-qN\",\"number\":N,\"type\":\"advanced_math\",\"prompt\":\"...\",\"answer_format\":\"student_produced\",\"correct_answer\":\"numeric_value\"}\nPlace SPR questions at the END of the array.\n\nFollow instructions for your type:\n\n[algebra] (MCQ + 2 SPR)\n- Topics: linear equations (1 var, 2 vars), systems of 2 linear equations, linear inequalities, absolute value\n- Mix word problems and abstract algebra. Use $...$ notation.\n- MCQ distractors: sign errors, arithmetic mistakes, setup errors, partial answers\n- SPR answers: clean integers or simple fractions\n\n[functions_analysis] (MCQ + 2 SPR)\n- Topics: function evaluation, domain/range, transformations (f(x)+k, f(x-h)), graphical analysis, composition f(g(x)), piecewise\n- At least one conceptual insight question (e.g., (p,q) and (q,p) on same graph)\n- Describe graphs in text. All computations yield clean results.\n\n[geometry_trigonometry] (MCQ + 2 SPR)\n- Topics: circle equations, Pythagorean theorem, special right triangles, coordinate geometry, trig ratios, volume/surface area, angle relationships\n- Optional \"visual\" field: {\"type\":\"coordinate_plane\",\"description\":\"...\"}\n- Radical answers OK (e.g., $5\\sqrt{13}$). Multi-step reasoning required.\n\n[statistics_probability] (MCQ + 2 SPR)\n- Topics: mean/median/mode, standard deviation (conceptual), probability (conditional from tables), data interpretation, scatter plots/line of best fit, sampling/bias\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Distractors: P(A|B) vs P(B|A) confusion, mean vs median, causal from correlational\n\n[advanced_math] (MCQ + 2 SPR)\n- Topics: quadratics (factoring, discriminant), polynomials (zeros, end behavior), exponentials (growth/decay), rational expressions, radical expressions, non-linear systems\n- Hardest questions on the test. Multi-step reasoning required. Elegant algebraic insight preferred.\n- All expressions must be factorable with integer coefficients. SPR answers are clean numbers.\n\nVERIFY EVERY ANSWER: Solve each problem step-by-step before finalizing correct_answer.\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "```json\n[\n {\n \"id\": \"sat11-math-m2-q23\",\n \"number\": 23,\n \"type\": \"advanced_math\",\n \"prompt\": \"The parabola with equation $y = x^2 - 5x + 11$ intersects the line with equation $y = 3x - 4$ at two points, $(x_1, y_1)$ and $(x_2, y_2)$. What is the value of $x_1 + x_2$?\",\n \"choices\": {\n \"A\": \"8\",\n \"B\": \"15\",\n \"C\": \"2\",\n \"D\": \"-2\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\"\n },\n {\n \"id\": \"sat11-math-m2-q24\",\n \"number\": 24,\n \"type\": \"advanced_math\",\n \"prompt\": \"The expression $\\\\frac{3x^2+5x-2}{x^2-4}$ can be written as $A + \\\\frac{B}{x-2}$, where $A$ and $B$ are constants and $x \\\\neq \\\\pm 2$. What is the value of $B$?\",\n \"choices\": {\n \"A\": \"3\",\n \"B\": \"5\",\n \"C\": \"8.5\",\n \"D\": \"4.5\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"D\"\n },\n {\n \"id\": \"sat11-math-m2-q25\",\n \"number\": 25,\n \"type\": \"advanced_math\",\n \"prompt\": \"For a function defined by $f(x) = c \\cdot p^x$, where $c$ and $p$ are positive constants, $f(1)=15$ and $f(3)=135$. What is the value of $p$?\",\n \"choices\": {\n \"A\": \"9\",\n \"B\": \"5\",\n \"C\": \"3\",\n \"D\": \"60\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m2-q26\",\n \"number\": 26,\n \"type\": \"advanced_math\",\n \"prompt\": \"If $x > 0$ and $2\\sqrt{x+2} = x+1$, what is the value of $x$?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": 2\n },\n {\n \"id\": \"sat11-math-m2-q27\",\n \"number\": 27,\n \"type\": \"advanced_math\",\n \"prompt\": \"In the $xy$-plane, the graphs of the functions $f(x) = -x^2+3$ and $g(x) = x^2+bx-1$ intersect at exactly one point. What is the value of $b$?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": 4\n }\n]\n```" }, "status": "ok" } ], "succeeded": 10 } }, "upstream": { "gen-rw": { "count": 14, "costUsd": 2.06721, "failed": 0, "items": [ { "costUsd": 0.084615, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Reading & Writing questions.\nType: vocabulary_in_context | Test: 11 | Unit: 1 | Start Q#: 1\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m1-qN\" where N is the sequential question number starting at 1.\nFor example, the first question in this batch has id \"sat11-rw-m1-q1\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m1-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m1-q1\",\"number\":1,\"type\":\"vocabulary_in_context\",\"passage\":\"In the decades following the publication of Keynes's General Theory, economists were reluctant to ______ the orthodoxy of classical market equilibrium outright. Instead, they tended to graft Keynesian insights onto existing models, preserving much of the older framework even as its foundational assumptions were being quietly eroded.\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"repudiate\",\"B\":\"disregard\",\"C\":\"misconstrue\",\"D\":\"lament\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m1-q2\",\"number\":2,\"type\":\"vocabulary_in_context\",\"passage\":\"Although Giorgio Vasari praised Pontormo's late frescoes for their technical daring, later critics found the figures' elongated limbs and unnatural palette deeply ______. To these observers, the Mannerist departure from High Renaissance proportion signaled not innovation but a troubling loss of classical composure.\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"prosaic\",\"B\":\"derivative\",\"C\":\"disconcerting\",\"D\":\"ostentatious\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m1-q3\",\"number\":3,\"type\":\"vocabulary_in_context\",\"passage\":\"Behavioral economists have long observed that consumers do not weigh gains and losses symmetrically; a loss of \\\\$100 tends to ______ the pleasure of a \\\\$100 gain. This asymmetry, known as loss aversion, helps explain why investors often hold losing stocks far longer than rational models predict.\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"eclipse\",\"B\":\"replicate\",\"C\":\"supplement\",\"D\":\"anticipate\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m1-q4\",\"number\":4,\"type\":\"vocabulary_in_context\",\"passage\":\"Because the faint signal from the distant quasar was nearly ______ by the glow of intervening galaxies, astronomers at the Keck Observatory employed adaptive optics to isolate its spectrum. Only after filtering out the foreground light could they confirm the object's redshift with any confidence.\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"amplified\",\"B\":\"refracted\",\"C\":\"subsumed\",\"D\":\"occluded\"},\"answer_format\":\"mcq\",\"correct_answer\":\"D\"},{\"id\":\"sat11-rw-m1-q5\",\"number\":5,\"type\":\"vocabulary_in_context\",\"passage\":\"Hannah Arendt argued that totalitarian regimes do not merely suppress dissent; they seek to ______ the very categories through which citizens distinguish truth from falsehood. Once this shared epistemic ground collapses, she warned, resistance itself becomes almost unintelligible.\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"codify\",\"B\":\"dissolve\",\"C\":\"arbitrate\",\"D\":\"venerate\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}]" }, { "costUsd": 0.25419, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 8 SAT Reading & Writing questions.\nType: text_comprehension | Test: 11 | Unit: 1 | Start Q#: 6\n\nMANDATORY: Your JSON array MUST contain exactly 8 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m1-qN\" where N is the sequential question number starting at 6.\nFor example, the first question in this batch has id \"sat11-rw-m1-q6\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m1-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 8 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m1-q6\",\"number\":6,\"type\":\"text_comprehension\",\"passage\":\"The following is adapted from Willa Cather's 1918 novel My Ántonia. The narrator recalls his arrival at his grandparents' Nebraska farm as a child.\\n\\nI had the sense of coming home to myself, and of having found out what a little circle man's experience is. For the first time, perhaps, since that night, I was dissolved into something complete and great. When it came to me, it came as naturally as sleep. The wind sang a sort of clear tune through the long grass, and I lay down among the pumpkins and did not try to think of anything.\",\"prompt\":\"Which choice best describes the narrator's state of mind in the passage?\",\"choices\":{\"A\":\"He feels a profound sense of peace that quiets his impulse toward reflection.\",\"B\":\"He is disoriented by the unfamiliar landscape and seeks refuge among the pumpkins.\",\"C\":\"He is eager to explore the farm but exhausted from his long journey.\",\"D\":\"He is nostalgic for a former home he has recently left behind.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m1-q7\",\"number\":7,\"type\":\"text_comprehension\",\"passage\":\"In a 2019 study, ecologist Mariana Altrichter examined the role of white-lipped peccaries in the Gran Chaco forests of Paraguay. Altrichter found that the peccaries, which travel in herds of up to three hundred, create wallows—shallow muddy depressions—that persist long after the herds have moved on. These wallows retain water during dry seasons and become refuges for frogs, insects, and small mammals that would otherwise perish. When peccary populations collapse due to hunting, the wallows dry up within a single season, and local biodiversity declines sharply.\",\"prompt\":\"Which choice best states the main idea of the text?\",\"choices\":{\"A\":\"Peccary herds in the Gran Chaco have grown in size as hunting pressures have decreased in recent decades.\",\"B\":\"Altrichter's 2019 study was the first to document the behavior of white-lipped peccaries in Paraguay.\",\"C\":\"The disappearance of peccaries threatens the frogs and insects that share their habitat in the Gran Chaco.\",\"D\":\"By creating wallows that sustain other species, peccaries function as ecosystem engineers whose loss reverberates through the Chaco food web.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"D\"},{\"id\":\"sat11-rw-m1-q8\",\"number\":8,\"type\":\"text_comprehension\",\"passage\":\"The following is adapted from a 2021 essay on the composer Florence Price. After Price's death in 1953, most of her manuscripts were stored in an abandoned summer home outside St. Anne, Illinois. In 2009, renovators preparing the house for demolition discovered trunks of her scores, including two violin concertos and the Fourth Symphony, works that had never been performed publicly. The recovery has prompted orchestras to reevaluate a composer who, during her lifetime, was celebrated in Chicago but largely ignored by the East Coast musical establishment.\",\"prompt\":\"The author's main purpose in the passage is to\",\"choices\":{\"A\":\"argue that Price's Fourth Symphony deserves a place in the standard orchestral repertoire.\",\"B\":\"describe a chance discovery that has reshaped the reception of a previously overlooked composer.\",\"C\":\"contrast the musical cultures of Chicago and the East Coast in the early twentieth century.\",\"D\":\"criticize the institutions that failed to preserve Price's manuscripts during her lifetime.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m1-q9\",\"number\":9,\"type\":\"text_comprehension\",\"passage\":\"In The Federalist No. 10 (1787), James Madison addressed the problem of \\\"faction\\\"—groups of citizens united by a common interest adverse to the rights of others or to the public good. Rather than attempting to eliminate factions by suppressing liberty or enforcing uniformity of opinion, Madison argued that the remedy lay in controlling their effects. A large republic, he reasoned, would encompass so many competing interests that no single faction could easily assemble a majority, and elected representatives would tend to refine the raw preferences of constituents into more deliberate judgments.\",\"prompt\":\"Based on the text, which choice best describes Madison's view of factions?\",\"choices\":{\"A\":\"Factions are an unavoidable consequence of liberty, and their dangers are best managed through the structural features of a large republic.\",\"B\":\"Factions arise chiefly from economic inequality, which representative government is well equipped to correct.\",\"C\":\"Factions pose little threat so long as elected representatives share the moral values of their constituents.\",\"D\":\"Factions should be suppressed through legal restrictions that preserve uniformity of opinion among citizens.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m1-q10\",\"number\":10,\"type\":\"text_comprehension\",\"passage\":\"The following is adapted from Edith Wharton's 1905 novel The House of Mirth. Lily Bart, a young socialite, has just encountered Lawrence Selden at a train station.\\n\\nSelden had in fact recognized her at once, and he now advanced toward her with a certain hesitancy. It was characteristic of her to express herself so quickly, and of him to measure his words before committing them to speech. He wondered, as he looked at her, how many years it would take her to learn what it had taken him so long to unlearn: that life, at its deepest, rewarded not the swift gesture but the patient observation.\",\"prompt\":\"Which choice best describes the function of the underlined portion (\\\"He wondered...observation\\\") in the text as a whole?\",\"choices\":{\"A\":\"It reveals Selden's irritation at Lily's impulsive manner and his desire to correct her.\",\"B\":\"It suggests that Selden has come to value reflection over spontaneity and senses that Lily has not yet done so.\",\"C\":\"It implies that Selden envies Lily's quickness and regrets his own deliberate temperament.\",\"D\":\"It indicates that Selden and Lily share a temperament that both have cultivated over many years.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m1-q11\",\"number\":11,\"type\":\"text_comprehension\",\"passage\":\"Astrophysicist Wendy Freedman has long maintained that the Hubble constant—a measure of the universe's expansion rate—cannot be pinned down without first resolving systematic errors in the distance ladder used to calibrate it. In a 2023 paper, Freedman and her collaborators measured distances to nearby galaxies using the tip of the red giant branch (TRGB), a technique that bypasses Cepheid variable stars. Their value of 69.8 km/s/Mpc falls between the higher figure derived from Cepheid-based measurements and the lower figure inferred from the cosmic microwave background, a discrepancy known as the \\\"Hubble tension.\\\"\",\"prompt\":\"Based on the text, what can most reasonably be inferred about the \\\"Hubble tension\\\"?\",\"choices\":{\"A\":\"Freedman's 2023 measurement definitively resolves the Hubble tension in favor of cosmic microwave background estimates.\",\"B\":\"The Hubble tension will likely be resolved once astronomers abandon the use of Cepheid variables as distance indicators.\",\"C\":\"Freedman's intermediate value suggests that unaccounted systematic errors, rather than new physics, may underlie the discrepancy.\",\"D\":\"The Hubble tension arises primarily because the cosmic microwave background cannot be observed at sufficient resolution.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m1-q12\",\"number\":12,\"type\":\"text_comprehension\",\"passage\":\"In studies of intertemporal choice, behavioral economists have long observed that people discount future rewards more steeply than classical models predict. Offered $50 today or $55 in a week, many subjects take the $50; offered $50 in a year or $55 in a year and a week, the same subjects wait. This pattern, termed \\\"hyperbolic discounting,\\\" is difficult to reconcile with the assumption that preferences remain stable across time. Richard Thaler and others have argued that the inconsistency reflects not irrationality but a conflict between a patient \\\"planner\\\" self and an impulsive \\\"doer\\\" self, each exerting influence at different moments.\",\"prompt\":\"Which choice best describes the overall structure of the text?\",\"choices\":{\"A\":\"It identifies a puzzling behavioral pattern and then summarizes one theoretical account offered to explain it.\",\"B\":\"It presents a longstanding economic theory and then describes experimental evidence that decisively refutes it.\",\"C\":\"It compares two competing schools of thought within behavioral economics and favors one over the other.\",\"D\":\"It traces the historical development of hyperbolic discounting from its origins to its current applications.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m1-q13\",\"number\":13,\"type\":\"text_comprehension\",\"passage\":\"Art historian Patricia Rubin has argued that the workshop of Andrea del Verrocchio, active in Florence during the 1470s, functioned less as a training ground for imitators than as a laboratory in which young artists—including the young Leonardo da Vinci—tested new techniques against one another. Surviving panels from the workshop often bear the hands of multiple contributors, and Rubin contends that Verrocchio deliberately assigned passages of the same painting to different apprentices so that their divergent solutions could be studied and compared.\",\"prompt\":\"Which choice best describes the author's tone in the passage?\",\"choices\":{\"A\":\"Skeptical of Rubin's interpretation and inclined to offer an alternative.\",\"B\":\"Neutral in summarizing Rubin's argument without endorsing or rejecting it.\",\"C\":\"Admiring of Verrocchio's business acumen in running a profitable workshop.\",\"D\":\"Dismissive of earlier scholarship on Renaissance apprenticeship practices.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}]" }, { "costUsd": 0.146565, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 4 SAT Reading & Writing questions.\nType: data_analysis | Test: 11 | Unit: 1 | Start Q#: 14\n\nMANDATORY: Your JSON array MUST contain exactly 4 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m1-qN\" where N is the sequential question number starting at 14.\nFor example, the first question in this batch has id \"sat11-rw-m1-q14\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m1-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 4 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m1-q14\",\"number\":14,\"type\":\"data_analysis\",\"passage\":\"Marine biologist Hina Watanabe investigated how water temperature affects the foraging efficiency of juvenile cuttlefish (Sepia officinalis). She measured the average number of successful prey captures per hour for cuttlefish housed in tanks set to four different temperatures, recording results across a six-week observation period.\",\"visual\":{\"type\":\"table\",\"title\":\"Average Prey Captures per Hour by Water Temperature\",\"headers\":[\"Water Temperature (°C)\",\"Average Captures per Hour\",\"Standard Deviation\"],\"rows\":[[\"14\",\"3.2\",\"0.6\"],[\"18\",\"5.8\",\"0.7\"],[\"22\",\"7.1\",\"0.5\"],[\"26\",\"4.4\",\"0.9\"]]},\"prompt\":\"Which statement about the relationship between water temperature and foraging efficiency is best supported by the data?\",\"choices\":{\"A\":\"Foraging efficiency increases steadily as water temperature rises across the entire tested range.\",\"B\":\"Foraging efficiency peaks at an intermediate temperature and declines at both lower and higher temperatures.\",\"C\":\"Foraging efficiency is highest at the coldest temperature tested and decreases as the water warms.\",\"D\":\"Foraging efficiency remains essentially constant regardless of changes in water temperature.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m1-q15\",\"number\":15,\"type\":\"data_analysis\",\"passage\":\"In a behavioral economics study, researcher Diego Alarcón asked participants to choose between receiving a smaller monetary reward immediately or a larger reward after a delay. He recorded the percentage of participants who chose the delayed option at each combination of delay length and reward size, hypothesizing that participants would tolerate longer delays only when the delayed reward was substantially larger than the immediate one.\",\"visual\":{\"type\":\"table\",\"title\":\"Percentage Choosing Delayed Reward over $20 Immediate Payment\",\"headers\":[\"Delay Length\",\"Delayed Reward $25\",\"Delayed Reward $50\",\"Delayed Reward $100\"],\"rows\":[[\"1 week\",\"62%\",\"88%\",\"95%\"],[\"1 month\",\"31%\",\"64%\",\"82%\"],[\"6 months\",\"9%\",\"28%\",\"57%\"]]},\"prompt\":\"Which statement about the data is most consistent with Alarcón's hypothesis?\",\"choices\":{\"A\":\"At every delay length, a larger delayed reward produced a higher percentage of participants willing to wait.\",\"B\":\"Participants always preferred the immediate $20 reward over any delayed reward offered in the study.\",\"C\":\"The percentage of participants choosing the delayed reward depended only on the delay length, not on the reward size.\",\"D\":\"Increasing the delayed reward from $25 to $100 had no measurable effect when the delay was 6 months.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m1-q16\",\"number\":16,\"type\":\"data_analysis\",\"passage\":\"Agronomist Felipe Ortega tested how nitrogen fertilizer application affects yield in a variety of winter wheat. He applied four different rates of nitrogen to otherwise identical plots and recorded the average grain yield, measured in bushels per acre, at the end of the growing season. Ortega claimed that beyond a certain rate, additional nitrogen produces only minimal gains in yield.\",\"visual\":{\"type\":\"table\",\"title\":\"Winter Wheat Yield by Nitrogen Application Rate\",\"headers\":[\"Nitrogen Applied (lb/acre)\",\"Average Yield (bushels/acre)\"],\"rows\":[[\"40\",\"52\"],[\"80\",\"71\"],[\"120\",\"83\"],[\"160\",\"85\"]]},\"prompt\":\"Which finding from the table best supports Ortega's claim?\",\"choices\":{\"A\":\"Increasing nitrogen from 40 to 80 lb/acre raised yield by 19 bushels per acre.\",\"B\":\"Increasing nitrogen from 80 to 120 lb/acre raised yield by 12 bushels per acre.\",\"C\":\"Increasing nitrogen from 120 to 160 lb/acre raised yield by only 2 bushels per acre.\",\"D\":\"The plot receiving 40 lb/acre produced the lowest yield in the study.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m1-q17\",\"number\":17,\"type\":\"data_analysis\",\"passage\":\"Sociologist Marta Kowalski surveyed adults in four metropolitan areas about their weekly volunteer hours and their reported sense of community belonging, scored on a 0-to-10 scale. She wanted to determine whether residents who volunteered more also tended to report a stronger connection to their neighborhood, while cautioning that her data could not establish a causal direction.\",\"visual\":{\"type\":\"table\",\"title\":\"Average Volunteer Hours and Belonging Scores by City\",\"headers\":[\"City\",\"Avg. Weekly Volunteer Hours\",\"Avg. Belonging Score (0-10)\"],\"rows\":[[\"Riverton\",\"1.2\",\"4.3\"],[\"Glenford\",\"2.5\",\"5.9\"],[\"Pinehill\",\"3.8\",\"7.2\"],[\"Westmere\",\"5.1\",\"8.4\"]]},\"prompt\":\"Which conclusion is most strongly supported by Kowalski's data, given her stated caution?\",\"choices\":{\"A\":\"Volunteering more hours each week causes residents to feel a stronger sense of community belonging.\",\"B\":\"Cities with higher average volunteer hours tend to have higher average belonging scores, but the data alone cannot show that one variable causes the other.\",\"C\":\"Residents in cities with low belonging scores would increase their belonging if they were required to volunteer more.\",\"D\":\"Belonging scores and volunteer hours appear to be unrelated across the four cities surveyed.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}]" }, { "costUsd": 0.11679, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 4 SAT Reading & Writing questions.\nType: claim_support | Test: 11 | Unit: 1 | Start Q#: 18\n\nMANDATORY: Your JSON array MUST contain exactly 4 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m1-qN\" where N is the sequential question number starting at 18.\nFor example, the first question in this batch has id \"sat11-rw-m1-q18\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m1-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 4 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"Archaeologist Dr. Marisela Ortega has proposed that the inhabitants of the pre-Columbian settlement of Tiwanaku engaged in long-distance trade with coastal communities as early as 500 CE, rather than relying solely on local resources as previously assumed.\",\"prompt\":\"Which finding, if true, would most directly support Dr. Ortega's hypothesis?\",\"choices\":{\"A\":\"Stone tools found at Tiwanaku exhibit craftsmanship techniques similar to those documented at other highland sites of the same period.\",\"B\":\"Ceramic vessels excavated from Tiwanaku contain residues of marine fish and mollusks that are not native to the highland region.\",\"C\":\"Agricultural terraces near Tiwanaku show evidence of sophisticated irrigation systems that could support a large population.\",\"D\":\"Burial sites at coastal communities contain ornamental objects whose stylistic features differ markedly from those at Tiwanaku.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m1-q19\",\"number\":19,\"type\":\"claim_support\",\"passage\":\"In her 2019 study of urban songbird populations, ornithologist Priya Chandran argued that house finches living in cities have adapted their vocalizations to compensate for low-frequency traffic noise, allowing them to communicate more effectively than their rural counterparts in noisy environments.\",\"prompt\":\"Which finding from Chandran's study would most directly support her claim?\",\"choices\":{\"A\":\"Urban house finches produce songs with higher minimum frequencies than rural house finches, and these songs travel farther when ambient traffic noise is present.\",\"B\":\"Rural house finches and urban house finches exhibit nearly identical body sizes, plumage patterns, and foraging behaviors across the study sites.\",\"C\":\"Urban house finches sing more frequently during early morning hours, when traffic volume in the surrounding neighborhoods tends to be at its lowest.\",\"D\":\"Rural house finch populations have remained stable over the past decade, while urban populations have fluctuated in response to habitat changes.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m1-q20\",\"number\":20,\"type\":\"claim_support\",\"passage\":\"A literary scholar contends that in Virginia Woolf's novel Mrs. Dalloway, the recurring sound of Big Ben functions not merely as a marker of time but as a mechanism that forcibly returns characters from their private reveries to the shared public world of London.\",\"prompt\":\"Which quotation from Mrs. Dalloway would most effectively illustrate the scholar's claim?\",\"choices\":{\"A\":\"\\\"For having lived in Westminster—how many years now? over twenty,—one feels even in the midst of the traffic, or waking at night, Clarissa was positive, a particular hush.\\\"\",\"B\":\"\\\"She had a perpetual sense, as she watched the taxi cabs, of being out, out, far out to sea and alone; she always had the feeling that it was very, very dangerous to live even one day.\\\"\",\"C\":\"\\\"The leaden circles dissolved in the air. Such fools we are, she thought, crossing Victoria Street, and immediately she was back in the noise of the present, hearing the omnibuses grind.\\\"\",\"D\":\"\\\"He had left a note for her, written in pencil, her favourite, saying he would be back; but the words had slipped from her mind like water through fingers.\\\"\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m1-q21\",\"number\":21,\"type\":\"claim_support\",\"passage\":\"Behavioral economist Tomás Reyes hypothesizes that consumers are more likely to complete an online purchase when the checkout process explicitly displays the number of items remaining in stock, because perceived scarcity triggers a sense of urgency that overrides hesitation about the purchase.\",\"prompt\":\"Which finding, if true, would most directly undermine Reyes's hypothesis?\",\"choices\":{\"A\":\"Shoppers presented with \\\"only 3 left in stock\\\" messaging report feeling more anxious during checkout than those who see no such messaging.\",\"B\":\"Websites displaying stock counts generate higher total revenue per visitor than websites that do not display such counts.\",\"C\":\"Completion rates at checkout remain statistically identical whether or not stock-count messaging appears, regardless of how low the count displayed.\",\"D\":\"Consumers who abandon their carts most often cite shipping costs, rather than product availability, as their primary reason for leaving.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"}]" }, { "costUsd": 0.158865, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Reading & Writing questions.\nType: standard_english_conventions | Test: 11 | Unit: 1 | Start Q#: 22\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m1-qN\" where N is the sequential question number starting at 22.\nFor example, the first question in this batch has id \"sat11-rw-m1-q22\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m1-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m1-q22\",\"number\":22,\"type\":\"standard_english_conventions\",\"passage\":\"The extensive collection of botanical illustrations housed at the Royal Horticultural Society's Lindley Library [underlined: are] considered among the finest examples of nineteenth-century scientific artistry in Europe.\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"is\",\"C\":\"were\",\"D\":\"have been\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m1-q23\",\"number\":23,\"type\":\"standard_english_conventions\",\"passage\":\"After reviewing quarterly filings submitted by several multinational firms, the auditor concluded that the [underlined: companies reports] contained irregularities warranting further investigation by regulators.\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"company's reports\",\"C\":\"companies' reports\",\"D\":\"companies reports'\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m1-q24\",\"number\":24,\"type\":\"standard_english_conventions\",\"passage\":\"The nebula NGC 6888, commonly known as the Crescent Nebula, formed approximately 250,000 years [underlined: ago, it was] shaped by powerful stellar winds emanating from a central Wolf-Rayet star.\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"ago, being\",\"C\":\"ago, and\",\"D\":\"ago; it was\"},\"answer_format\":\"mcq\",\"correct_answer\":\"D\"},{\"id\":\"sat11-rw-m1-q25\",\"number\":25,\"type\":\"standard_english_conventions\",\"passage\":\"By the time archaeologist Howard Carter reached the inner chamber of Tutankhamun's tomb in 1923, ancient looters [underlined: had already removed] a small portion of the burial goods, though the bulk of the treasure remained undisturbed.\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"have already removed\",\"C\":\"already remove\",\"D\":\"are already removing\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m1-q26\",\"number\":26,\"type\":\"standard_english_conventions\",\"passage\":\"In her report to the city council, urban planner Yuki Tanaka recommended that the revised zoning policy emphasize transparency, [underlined: encourages] community accountability, and promote equitable access to public resources.\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"encourage\",\"C\":\"encouraging\",\"D\":\"had encouraged\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}]" }, { "costUsd": 0.08049, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 4 SAT Reading & Writing questions.\nType: transitions | Test: 11 | Unit: 1 | Start Q#: 27\n\nMANDATORY: Your JSON array MUST contain exactly 4 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m1-qN\" where N is the sequential question number starting at 27.\nFor example, the first question in this batch has id \"sat11-rw-m1-q27\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m1-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 4 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m1-q27\",\"number\":27,\"type\":\"transitions\",\"passage\":\"The Antikythera mechanism, an ancient Greek device recovered from a shipwreck in 1901, was long assumed to be a simple astronomical clock. ______ recent X-ray tomography has revealed more than thirty interlocking bronze gears capable of predicting eclipses and tracking the irregular motion of the moon, suggesting a level of mechanical sophistication previously thought impossible for the second century BCE.\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"However,\",\"B\":\"Likewise,\",\"C\":\"Consequently,\",\"D\":\"For instance,\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m1-q28\",\"number\":28,\"type\":\"transitions\",\"passage\":\"Economist Elinor Ostrom challenged the prevailing view that shared resources such as fisheries and grazing lands inevitably suffer from overuse. Through decades of fieldwork in Nepal, Kenya, and the Swiss Alps, she documented communities that had sustainably managed common pools for centuries. ______ she identified eight design principles, including clearly defined boundaries and graduated sanctions, that reliably predict which communal arrangements endure.\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"Nevertheless,\",\"B\":\"From these case studies,\",\"C\":\"In contrast,\",\"D\":\"Conversely,\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m1-q29\",\"number\":29,\"type\":\"transitions\",\"passage\":\"The painter Artemisia Gentileschi worked in a profession almost entirely closed to women in seventeenth-century Italy, and her early biographers often framed her career as a curiosity rather than an artistic achievement. Art historian Mary Garrard, writing in the 1980s, argued that this framing obscured the formal rigor of Gentileschi's compositions. ______ her influential monograph reoriented scholarly attention toward the technical choices visible in works such as Judith Slaying Holofernes.\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"Nonetheless,\",\"B\":\"Similarly,\",\"C\":\"Accordingly,\",\"D\":\"In particular,\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m1-q30\",\"number\":30,\"type\":\"transitions\",\"passage\":\"Cuttlefish possess chromatophores, pigment-filled sacs in their skin that expand and contract to produce rapid color changes. These displays are used for camouflage, mating signals, and territorial warnings. ______ cuttlefish are colorblind: their single type of photoreceptor cannot distinguish wavelengths, meaning they coordinate elaborate visual performances they themselves may not fully perceive.\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"Moreover,\",\"B\":\"Therefore,\",\"C\":\"Paradoxically,\",\"D\":\"Meanwhile,\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"}]" }, { "costUsd": 0.11544, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 3 SAT Reading & Writing questions.\nType: rhetorical_synthesis | Test: 11 | Unit: 1 | Start Q#: 31\n\nMANDATORY: Your JSON array MUST contain exactly 3 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m1-qN\" where N is the sequential question number starting at 31.\nFor example, the first question in this batch has id \"sat11-rw-m1-q31\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m1-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 3 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m1-q31\",\"number\":31,\"type\":\"rhetorical_synthesis\",\"passage\":\"While researching the lyrebird, a student has taken these notes:\\n- The superb lyrebird (Menura novaehollandiae) is native to southeastern Australia.\\n- Male lyrebirds are renowned for their vocal mimicry.\\n- A 2021 study by Anastasia Dalziell found that male lyrebirds mimic the alarm calls of multiple bird species during courtship.\\n- Dalziell concluded that this mimicry may deceive females into remaining nearby, mistaking the sounds for a genuine predator threat.\",\"prompt\":\"The student wants to present the conclusion of Dalziell's 2021 study. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"Native to southeastern Australia, the superb lyrebird (Menura novaehollandiae) is famous for the vocal mimicry of its males.\",\"B\":\"In a 2021 study, Anastasia Dalziell observed that male superb lyrebirds mimic the alarm calls of multiple bird species during courtship.\",\"C\":\"Dalziell concluded that male lyrebirds' courtship mimicry may deceive females into staying nearby by simulating a predator threat.\",\"D\":\"Male superb lyrebirds, native to southeastern Australia, are capable of producing the alarm calls of several other bird species.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m1-q32\",\"number\":32,\"type\":\"rhetorical_synthesis\",\"passage\":\"While researching microplastics in marine ecosystems, a student has taken these notes:\\n- Microplastics are plastic fragments smaller than 5 millimeters.\\n- A 2022 survey of the North Atlantic found an average of 1.2 microplastic particles per cubic meter of seawater.\\n- A 2023 survey of the Mediterranean Sea found an average of 4.7 microplastic particles per cubic meter of seawater.\\n- Both surveys used identical sampling methods.\",\"prompt\":\"The student wants to emphasize a difference between the two surveys' findings. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"Microplastics, defined as plastic fragments smaller than 5 millimeters, are now found throughout the world's oceans, including the North Atlantic and the Mediterranean.\",\"B\":\"Although both surveys used identical sampling methods, the 2023 Mediterranean survey detected 4.7 microplastic particles per cubic meter, nearly four times the 1.2 particles per cubic meter recorded in the 2022 North Atlantic survey.\",\"C\":\"Two recent surveys, conducted in 2022 and 2023, measured the concentration of microplastic particles in samples drawn from the North Atlantic and the Mediterranean.\",\"D\":\"Microplastic particles—plastic fragments smaller than 5 millimeters—were detected in both the North Atlantic in 2022 and the Mediterranean in 2023.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m1-q33\",\"number\":33,\"type\":\"rhetorical_synthesis\",\"passage\":\"While researching the painter Artemisia Gentileschi, a student has taken these notes:\\n- Artemisia Gentileschi (1593–1656) was an Italian Baroque painter.\\n- She trained in the workshop of her father, Orazio Gentileschi.\\n- In 1616, she became the first woman admitted to the Accademia delle Arti del Disegno in Florence.\\n- Her paintings frequently depicted heroic women from biblical and mythological narratives.\",\"prompt\":\"The student wants to introduce Artemisia Gentileschi to an audience unfamiliar with her. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"In 1616, she became the first woman admitted to the Accademia delle Arti del Disegno in Florence.\",\"B\":\"Trained in her father Orazio's workshop, she frequently depicted heroic women from biblical and mythological narratives.\",\"C\":\"Artemisia Gentileschi (1593–1656) was an Italian Baroque painter known for her depictions of heroic women from biblical and mythological narratives.\",\"D\":\"Artemisia Gentileschi's paintings, which often featured biblical and mythological heroines, reflected the techniques she learned in her father's workshop.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"}]" }, { "costUsd": 0.08679, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Reading & Writing questions.\nType: vocabulary_in_context | Test: 11 | Unit: 2 | Start Q#: 1\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m2-qN\" where N is the sequential question number starting at 1.\nFor example, the first question in this batch has id \"sat11-rw-m2-q1\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m2-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m2-q1\",\"number\":1,\"type\":\"vocabulary_in_context\",\"passage\":\"In behavioral economics, researchers often find that small contextual cues can ______ decisions that individuals believe are made on purely rational grounds. A series of experiments by Thaler and Sunstein demonstrated that default options on retirement forms, though seemingly trivial, shifted enrollment rates by more than forty percent.\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"dictate\",\"B\":\"inflect\",\"C\":\"fabricate\",\"D\":\"commandeer\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q2\",\"number\":2,\"type\":\"vocabulary_in_context\",\"passage\":\"Although the fresco cycle in the Brancacci Chapel is now attributed largely to Masaccio, earlier attributions were often ______: nineteenth-century connoisseurs assigned individual panels to Masolino or Filippino Lippi on the basis of stylistic hunches that modern pigment analysis has since overturned.\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"conjectural\",\"B\":\"duplicitous\",\"C\":\"indifferent\",\"D\":\"exhaustive\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m2-q3\",\"number\":3,\"type\":\"vocabulary_in_context\",\"passage\":\"The protein chaperone Hsp90 does not merely fold its client molecules; it also ______ their activity by holding them in metastable conformations until a cellular signal permits full maturation. This capacity to pause and then release function is central to how eukaryotic cells cope with environmental stress.\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"abrogates\",\"B\":\"amplifies\",\"C\":\"modulates\",\"D\":\"misconstrues\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m2-q4\",\"number\":4,\"type\":\"vocabulary_in_context\",\"passage\":\"In $\\\\textit{The Social Contract}$, Rousseau argues that legitimate political authority cannot be ______ from any natural hierarchy; rather, it must arise from the free consent of equals who agree to be bound by a general will. This claim marked a decisive break from earlier theorists who grounded sovereignty in paternal or divine precedence.\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"rescinded\",\"B\":\"absolved\",\"C\":\"withheld\",\"D\":\"derived\"},\"answer_format\":\"mcq\",\"correct_answer\":\"D\"},{\"id\":\"sat11-rw-m2-q5\",\"number\":5,\"type\":\"vocabulary_in_context\",\"passage\":\"Observations from the James Webb Space Telescope have begun to ______ long-standing models of galaxy formation: massive, well-structured galaxies appear at redshifts so high that, under prevailing theories, they should not yet have had time to assemble. Cosmologists are now revisiting assumptions about the efficiency of early star formation.\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"unsettle\",\"B\":\"annihilate\",\"C\":\"ratify\",\"D\":\"postpone\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"}]" }, { "costUsd": 0.24009, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 8 SAT Reading & Writing questions.\nType: text_comprehension | Test: 11 | Unit: 2 | Start Q#: 6\n\nMANDATORY: Your JSON array MUST contain exactly 8 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m2-qN\" where N is the sequential question number starting at 6.\nFor example, the first question in this batch has id \"sat11-rw-m2-q6\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m2-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 8 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m2-q6\",\"number\":6,\"type\":\"text_comprehension\",\"passage\":\"In her 1862 letters, Emily Dickinson confided to her mentor Thomas Wentworth Higginson that she avoided publication because she feared the 'auction' of the mind. Yet she circulated dozens of poems privately, sewing them into small hand-bound booklets she called fascicles. Scholars have argued that these fascicles constitute a form of publication in themselves, one that preserved authorial control while still inviting a select readership.\",\"prompt\":\"Which choice best states the main idea of the text?\",\"choices\":{\"A\":\"Dickinson's refusal to publish her poems reflected a broader Victorian discomfort with female authorship.\",\"B\":\"Dickinson's fascicles can be understood as an alternative mode of publication rather than a retreat from readership.\",\"C\":\"Dickinson's correspondence with Higginson was the primary vehicle through which her poetry reached the public.\",\"D\":\"Dickinson's hand-bound booklets were technically inferior to commercial print but artistically superior.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q7\",\"number\":7,\"type\":\"text_comprehension\",\"passage\":\"The following is adapted from a 2019 article in a marine biology journal. Sea otters (Enhydra lutris) are often described as a keystone species because of their outsized effect on kelp-forest ecosystems. By preying on sea urchins, otters prevent the urchins from overgrazing kelp, which in turn shelters dozens of fish and invertebrate species. Recent surveys along the central California coast, however, suggest that the relationship is more conditional than once assumed: in bays where pycnopodia sea stars—another urchin predator—have collapsed, otters alone cannot suppress urchin populations enough to protect kelp.\",\"prompt\":\"What is the main purpose of the text?\",\"choices\":{\"A\":\"To argue that sea otters have been incorrectly classified as a keystone species in California kelp forests.\",\"B\":\"To describe a predator-prey relationship that has remained stable despite recent ecological disturbances.\",\"C\":\"To qualify a widely held claim about sea otters by presenting evidence that their ecological role depends on context.\",\"D\":\"To compare the hunting behaviors of sea otters and pycnopodia sea stars in protecting kelp forests.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m2-q8\",\"number\":8,\"type\":\"text_comprehension\",\"passage\":\"The following is adapted from Willa Cather's 1918 novel My Ántonia. The narrator Jim Burden recalls his grandmother's garden. I sat down in the middle of the garden, where snakes could scarcely approach unseen, and leaned my back against a warm yellow pumpkin. There were some ground-cherry bushes growing along the furrows, full of fruit. I turned back the papery triangular sheaths that protected the berries and ate a few. All about me giant grasshoppers, twice as big as any I had ever seen, were doing acrobatic feats among the dried vines. The gophers scurried up and down the ploughed ground.\",\"prompt\":\"Which choice best describes the function of the underlined details (the grasshoppers, gophers, and pumpkin) in the passage?\",\"choices\":{\"A\":\"They suggest that the garden has fallen into a state of neglect and disorder.\",\"B\":\"They establish Jim's detailed, sensory immersion in a landscape that feels abundant and alive.\",\"C\":\"They foreshadow a later conflict between Jim and the natural world of the prairie.\",\"D\":\"They contrast Jim's childhood memories with the harsher realities of farm labor.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q9\",\"number\":9,\"type\":\"text_comprehension\",\"passage\":\"Economist Elinor Ostrom, awarded the Nobel Prize in 2009, challenged the then-dominant assumption that shared resources—fisheries, grazing lands, irrigation systems—were inevitably overexploited unless governed by either private ownership or centralized state control. Drawing on fieldwork in villages from Nepal to Maine, she documented communities that had sustainably managed common resources for centuries through locally devised rules and mutual monitoring. Her work did not dispute that commons can collapse; rather, it identified the institutional conditions under which they do not.\",\"prompt\":\"Based on the text, how did Ostrom's research most likely reshape scholarly thinking about common resources?\",\"choices\":{\"A\":\"It demonstrated that privatization is the only reliable safeguard against resource depletion.\",\"B\":\"It showed that state regulation and private ownership are equally effective across all contexts.\",\"C\":\"It revealed a third pathway to sustainability that had been largely overlooked by earlier theorists.\",\"D\":\"It established that resource collapse is primarily caused by a failure of centralized enforcement.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m2-q10\",\"number\":10,\"type\":\"text_comprehension\",\"passage\":\"The following is adapted from an essay on Baroque painting. Caravaggio's 'The Calling of Saint Matthew' (1599–1600) resolves its dramatic tension through a single diagonal shaft of light that cuts across the shadowed tavern. Where earlier Mannerist painters crowded the picture plane with writhing figures and competing gestures, Caravaggio does the opposite: he empties the scene of ornament so that the viewer's eye has nowhere to travel but along the beam of light. What might in lesser hands have become theatrical instead becomes inevitable.\",\"prompt\":\"As used in the final sentence, 'inevitable' most nearly suggests that the composition feels\",\"choices\":{\"A\":\"oppressive and overburdened by symbolic weight.\",\"B\":\"compositionally necessary rather than merely staged.\",\"C\":\"derivative of earlier Mannerist conventions.\",\"D\":\"ambiguous in its spiritual implications.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q11\",\"number\":11,\"type\":\"text_comprehension\",\"passage\":\"A 2021 study of urban foxes in London by ecologist Kevin Parnell reported that foxes living within the M25 orbital motorway had, over roughly seventy generations, developed shorter, broader snouts and smaller brain cases than their rural counterparts. Parnell argued that these changes parallel the cranial shifts associated with domestication in dogs and cats, even though no human has deliberately bred the foxes. He suggested that the sustained proximity to humans, and the reduced selective pressure for wariness, may be producing a form of self-domestication.\",\"prompt\":\"Which choice best describes the structure of the text?\",\"choices\":{\"A\":\"A long-standing scientific theory is presented and then rejected in light of new genetic evidence.\",\"B\":\"A researcher's findings are summarized, followed by the interpretive framework he proposes to explain them.\",\"C\":\"Two competing explanations for an observed phenomenon are compared and one is endorsed.\",\"D\":\"An anecdotal observation is generalized into a claim about urban wildlife across Europe.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q12\",\"number\":12,\"type\":\"text_comprehension\",\"passage\":\"Historian Natalie Zemon Davis, writing about sixteenth-century France, once remarked that the archival record is 'not a mirror but a sieve.' What reaches the historian, she argued, has already been filtered by the interests of those who kept records: notaries, tax collectors, church officials. A peasant woman's dispute over a dowry survives only because a court clerk found it worth transcribing. The historian's task, then, is less to reconstruct a vanished totality than to read attentively against the grain of the sources that happen to remain.\",\"prompt\":\"Which choice best states the central claim of the text?\",\"choices\":{\"A\":\"Sixteenth-century French archives are too incomplete to support reliable historical scholarship.\",\"B\":\"Historians should prioritize the recovery of voices belonging to peasant women over those of clerks.\",\"C\":\"Because archival sources reflect the priorities of their creators, historians must interpret them critically rather than treating them as transparent windows onto the past.\",\"D\":\"The metaphor of the sieve is more accurate than the metaphor of the mirror in describing any form of historical evidence.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m2-q13\",\"number\":13,\"type\":\"text_comprehension\",\"passage\":\"In studies of infant cognition, researchers often rely on the 'violation of expectation' paradigm: infants are shown an event that either conforms to or breaks a physical rule, such as an object passing through a solid wall. Infants who look longer at the impossible event are presumed to have expected the possible one. Critics have noted, however, that longer looking times could reflect any number of reactions—surprise, confusion, visual novelty—and that inferring specific conceptual knowledge from gaze alone may credit infants with more than the data warrant.\",\"prompt\":\"Which choice best describes the author's attitude toward the violation-of-expectation paradigm?\",\"choices\":{\"A\":\"Enthusiastic endorsement of its power to reveal innate knowledge.\",\"B\":\"Cautious skepticism about the specificity of the inferences it supports.\",\"C\":\"Dismissive rejection of its value to developmental research.\",\"D\":\"Neutral description without evaluative commentary.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}]" }, { "costUsd": 0.344415, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 4 SAT Reading & Writing questions.\nType: data_analysis | Test: 11 | Unit: 2 | Start Q#: 14\n\nMANDATORY: Your JSON array MUST contain exactly 4 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m2-qN\" where N is the sequential question number starting at 14.\nFor example, the first question in this batch has id \"sat11-rw-m2-q14\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m2-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 4 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m2-q14\",\"number\":14,\"type\":\"data_analysis\",\"passage\":\"A team of plant physiologists measured the weekly growth of Boston fern specimens (Nephrolepis exaltata) cultivated under five different light intensities, holding temperature, humidity, and nutrients constant. A student summarizing the experiment writes: \\\"Although the fern responds strongly to modest increases in light at low intensities, ______\\\"\",\"visual\":{\"type\":\"table\",\"title\":\"Boston Fern Growth Rate by Light Intensity\",\"headers\":[\"Light Intensity (μmol/m²/s)\",\"Growth Rate (mm/week)\"],\"rows\":[[\"100\",\"4.2\"],[\"200\",\"7.8\"],[\"400\",\"11.5\"],[\"800\",\"12.9\"],[\"1600\",\"13.1\"]]},\"prompt\":\"Which choice most effectively uses data from the table to complete the statement?\",\"choices\":{\"A\":\"growth continues to rise in direct proportion to light intensity as it doubles from 400 to 1600 $\\\\mu mol/m^2/s$.\",\"B\":\"further increases in intensity beyond 400 $\\\\mu mol/m^2/s$ yield progressively smaller gains in growth rate.\",\"C\":\"growth rate declines sharply once light intensity exceeds 800 $\\\\mu mol/m^2/s$.\",\"D\":\"the relationship between light intensity and weekly growth is not evident in the measurements.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q15\",\"number\":15,\"type\":\"data_analysis\",\"passage\":\"In a behavioral-economics field study conducted at a Dutch supermarket, researchers varied the number of jam flavors displayed on a tasting booth and recorded both the percentage of passing shoppers who made a purchase and the percentage of purchasers who later reported being satisfied with their selection. A student summarizing the results claims that presenting more choices consistently reduces both the likelihood that a shopper will buy and the shopper's satisfaction with what was chosen.\",\"visual\":{\"type\":\"table\",\"title\":\"Purchase and Satisfaction Rates by Number of Jam Options\",\"headers\":[\"Number of Options\",\"Shoppers Who Purchased (%)\",\"Purchasers Satisfied with Choice (%)\"],\"rows\":[[\"6\",\"30\",\"84\"],[\"12\",\"22\",\"72\"],[\"18\",\"15\",\"64\"],[\"24\",\"10\",\"55\"]]},\"prompt\":\"Which choice most effectively uses data from the table to support the student's claim?\",\"choices\":{\"A\":\"At 6 options, 84% of purchasers reported satisfaction, while at 24 options only 10% of shoppers made a purchase.\",\"B\":\"As the number of options rose from 6 to 24, both the purchase rate and the satisfaction rate increased steadily.\",\"C\":\"Satisfaction rates among purchasers remained effectively constant regardless of the number of options displayed.\",\"D\":\"As the number of options rose from 6 to 24, the purchase rate fell from 30% to 10% and the satisfaction rate fell from 84% to 55%.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"D\"},{\"id\":\"sat11-rw-m2-q16\",\"number\":16,\"type\":\"data_analysis\",\"passage\":\"Astronomers detect exoplanets primarily through two techniques: the radial velocity (RV) method, which measures the stellar wobble induced by an orbiting planet, and transit photometry (TP), which measures periodic dimming as a planet crosses its star. A recent review tabulated confirmed detections by each method across four planetary mass bins. An astronomy student writes: \\\"Although the radial velocity technique is particularly effective for identifying massive planets, transit photometry has proven far more productive for detecting smaller worlds.\\\"\",\"visual\":{\"type\":\"table\",\"title\":\"Confirmed Exoplanet Detections by Method and Mass\",\"headers\":[\"Planet Mass (Earth masses)\",\"RV Method\",\"TP Method\"],\"rows\":[[\"< 1\",\"12\",\"47\"],[\"1–10\",\"89\",\"203\"],[\"10–100\",\"156\",\"118\"],[\"> 100\",\"74\",\"39\"]]},\"prompt\":\"Which choice most effectively uses data from the table to support the student's claim?\",\"choices\":{\"A\":\"For planets below 10 Earth masses, transit photometry confirmed more than twice as many detections as radial velocity (250 vs. 101), whereas for planets above 100 Earth masses, radial velocity confirmed nearly twice as many as transit photometry (74 vs. 39).\",\"B\":\"Radial velocity detected 12 planets with masses below 1 Earth mass, while transit photometry detected 39 planets with masses above 100 Earth masses.\",\"C\":\"Transit photometry produced more confirmed detections than radial velocity in every mass range listed in the table.\",\"D\":\"The radial velocity method has failed to confirm any planets with masses below 1 Earth mass.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m2-q17\",\"number\":17,\"type\":\"data_analysis\",\"passage\":\"An economist at the OECD compared the tertiary-education wage premium—the percentage by which workers with a university degree out-earn otherwise comparable workers whose schooling ended at the secondary level—for two age cohorts (25–34 and 55–64) in four member countries. A student summarizing the findings claims that in every country surveyed the wage premium conferred by a tertiary degree grows larger over the course of a worker's career.\",\"visual\":{\"type\":\"table\",\"title\":\"Tertiary-Education Wage Premium by Country and Age Cohort (%)\",\"headers\":[\"Country\",\"Ages 25–34\",\"Ages 55–64\"],\"rows\":[[\"Germany\",\"38\",\"76\"],[\"South Korea\",\"32\",\"58\"],[\"United States\",\"57\",\"92\"],[\"Italy\",\"28\",\"49\"]]},\"prompt\":\"Which choice most effectively uses data from the table to support the student's claim?\",\"choices\":{\"A\":\"The United States displayed the highest wage premium among workers aged 55–64, at 92 percent.\",\"B\":\"The wage premium in Italy was smaller than that in Germany for both age cohorts.\",\"C\":\"In each of the four countries, the wage premium for workers aged 55–64 was substantially higher than the premium for workers aged 25–34.\",\"D\":\"South Korea showed a smaller premium than Germany for younger workers but a comparable premium for older workers.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"}]" }, { "costUsd": 0.13599, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 4 SAT Reading & Writing questions.\nType: claim_support | Test: 11 | Unit: 2 | Start Q#: 18\n\nMANDATORY: Your JSON array MUST contain exactly 4 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m2-qN\" where N is the sequential question number starting at 18.\nFor example, the first question in this batch has id \"sat11-rw-m2-q18\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m2-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 4 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m2-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A team of marine biologists studying the Pacific bluefin tuna (Thunnus orynnalis) hypothesized that the species' ability to maintain a body temperature significantly warmer than the surrounding seawater allows it to forage efficiently in colder, deeper waters where prey density is higher. To test this, the team tagged 84 individuals with biologgers that recorded muscle temperature, ambient water temperature, and depth over a twelve-month period.\",\"prompt\":\"Which finding from the study, if true, would most directly support the researchers' hypothesis?\",\"choices\":{\"A\":\"Tagged tuna spent approximately equal proportions of time in surface waters and in deeper waters, regardless of the thermal gradient between their muscle tissue and the surrounding environment.\",\"B\":\"Individuals whose muscle temperatures exceeded ambient water temperatures by the largest margins executed significantly more feeding dives into cold, prey-rich waters below 400 meters than did individuals with smaller thermal differentials.\",\"C\":\"The average muscle temperature of tagged tuna was roughly 5 degrees Celsius higher than the ambient water temperature across all depths sampled during the study period.\",\"D\":\"Tuna that remained primarily in warm surface waters exhibited higher overall growth rates than those that frequently descended into colder, deeper zones where prey were more abundant.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q19\",\"number\":19,\"type\":\"claim_support\",\"passage\":\"In her 2019 study of Florentine workshop practice, art historian Marta Ricci argues that Andrea del Verrocchio's studio in the 1470s operated less as a conventional master-apprentice hierarchy and more as a collaborative environment in which senior apprentices were entrusted with substantive portions of commissioned works, including passages that bore the master's signature.\",\"prompt\":\"Which finding, if true, would most directly support Ricci's argument?\",\"choices\":{\"A\":\"Surviving account books from Verrocchio's workshop show that the master personally received payment for all commissions completed between 1470 and 1478.\",\"B\":\"Technical analysis of Verrocchio's Baptism of Christ reveals that the angel on the left and portions of the landscape were executed by the young Leonardo da Vinci using a distinctly different oil-glaze technique.\",\"C\":\"Contemporary Florentine guild regulations required that any panel painting bearing a master's signature be certified as having passed inspection by the workshop's head artisan.\",\"D\":\"Verrocchio is documented to have taken on an unusually large number of apprentices during the 1470s, which scholars have attributed to growing demand for sculptural commissions in the city.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q20\",\"number\":20,\"type\":\"claim_support\",\"passage\":\"In his essay collection The Weight of Ordinary Things (2021), critic Daniel Oyelaran contends that the short stories of Yiyun Li resist conventional emotional payoffs: rather than building toward catharsis, her narratives accumulate small, unresolved observations that leave the reader in a state of productive unease.\",\"prompt\":\"Which quotation from a Yiyun Li story would most effectively illustrate Oyelaran's claim?\",\"choices\":{\"A\":\"\\\"When the letter finally arrived, Mei wept openly in the kitchen, and for the first time in years she felt something loosen in her chest, as though a door she had long kept shut had swung wide.\\\"\",\"B\":\"\\\"He noticed the faint stain on the tablecloth, the way his sister's hand hovered near her teacup without lifting it, the silence that was not quite a silence; he said nothing, and neither did she, and eventually the waiter came.\\\"\",\"C\":\"\\\"The villain's confession, delivered in the final pages with theatrical flourish, resolved every question the novel had raised and sent readers home satisfied that justice had at last been served.\\\"\",\"D\":\"\\\"Anger rose in him with sudden clarity, and he understood, as if for the first time, exactly what he must do and why; the path ahead shone bright before him.\\\"\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q21\",\"number\":21,\"type\":\"claim_support\",\"passage\":\"An economist studying regional labor markets claims that increases in a state's minimum wage lead to measurable improvements in worker retention at low-wage firms because the higher wage reduces the incentive for employees to search for better-paying positions elsewhere.\",\"prompt\":\"Which finding, if true, would most directly support the economist's claim?\",\"choices\":{\"A\":\"Following a state-level minimum wage increase, the average voluntary quit rate among workers at low-wage firms in that state declined by 18 percent relative to neighboring states that did not raise their minimum wage.\",\"B\":\"States that raised their minimum wage experienced a slight decline in the total number of low-wage jobs available during the year following the policy change.\",\"C\":\"Surveys indicate that workers who earn above their state's minimum wage generally report higher levels of job satisfaction than workers who earn at or near the minimum.\",\"D\":\"After the minimum wage increase took effect, consumer prices at low-wage firms in affected states rose modestly, partially offsetting the real wage gains for workers.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"}]" }, { "costUsd": 0.095265, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Reading & Writing questions.\nType: standard_english_conventions | Test: 11 | Unit: 2 | Start Q#: 22\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m2-qN\" where N is the sequential question number starting at 22.\nFor example, the first question in this batch has id \"sat11-rw-m2-q22\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m2-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m2-q22\",\"number\":22,\"type\":\"standard_english_conventions\",\"passage\":\"In 1977, marine biologists aboard the submersible Alvin discovered hydrothermal vents near the Galápagos Rift. The communities of tube worms, clams, and crabs clustered around these vents [underlined: thrives] in an environment once thought incapable of supporting life.\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"thrive\",\"C\":\"has thrived\",\"D\":\"thriving\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q23\",\"number\":23,\"type\":\"standard_english_conventions\",\"passage\":\"The Ghanaian architect David Adjaye is known for designs that engage with cultural memory. His firm's most celebrated project, the Smithsonian National Museum of African American History and [underlined: Culture, opened] in 2016 on the National Mall in Washington, D.C.\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"Culture opened,\",\"C\":\"Culture; opened\",\"D\":\"Culture opened\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m2-q24\",\"number\":24,\"type\":\"standard_english_conventions\",\"passage\":\"Paleontologist Mary Anning, who spent much of her life collecting fossils along the cliffs of Lyme Regis, identified the first complete ichthyosaur skeleton when she was only twelve years old. [underlined: Despite her remarkable contributions to geology, she] was denied membership in the Geological Society of London because of her gender.\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"Despite her remarkable contributions to geology she\",\"C\":\"Despite her remarkable contributions to geology; she\",\"D\":\"Despite her, remarkable contributions to geology, she\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m2-q25\",\"number\":25,\"type\":\"standard_english_conventions\",\"passage\":\"Economist Elinor Ostrom challenged the prevailing view that shared resources inevitably collapse under private exploitation. Drawing on fieldwork in Nepal, Kenya, and Switzerland, [underlined: communities that manage fisheries, forests, and irrigation systems successfully were documented by Ostrom].\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"fisheries, forests, and irrigation systems were documented as successfully managed by communities in Ostrom's work\",\"C\":\"Ostrom documented communities that successfully manage fisheries, forests, and irrigation systems\",\"D\":\"there was documentation by Ostrom of communities successfully managing fisheries, forests, and irrigation systems\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m2-q26\",\"number\":26,\"type\":\"standard_english_conventions\",\"passage\":\"The Voyager 1 spacecraft, launched in 1977, crossed into interstellar space in 2012. Although its instruments now return data from more than 15 billion miles away, [underlined: their] plutonium power source is expected to fail sometime after 2025.\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"it's\",\"C\":\"its\",\"D\":\"they're\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"}]" }, { "costUsd": 0.08484, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 4 SAT Reading & Writing questions.\nType: transitions | Test: 11 | Unit: 2 | Start Q#: 27\n\nMANDATORY: Your JSON array MUST contain exactly 4 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m2-qN\" where N is the sequential question number starting at 27.\nFor example, the first question in this batch has id \"sat11-rw-m2-q27\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m2-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 4 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m2-q27\",\"number\":27,\"type\":\"transitions\",\"passage\":\"The rediscovery of Sofonisba Anguissola in the late twentieth century prompted museums to reattribute several sixteenth-century portraits long credited to male contemporaries. Curators initially hesitated, citing a lack of signed works by the artist. ______ subsequent archival research uncovered letters from the Spanish court confirming her authorship of numerous royal portraits, including ones previously ascribed to Alonso Sánchez Coello.\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"Likewise,\",\"B\":\"However,\",\"C\":\"Therefore,\",\"D\":\"For instance,\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q28\",\"number\":28,\"type\":\"transitions\",\"passage\":\"Behavioral economist Sendhil Mullainathan has argued that scarcity imposes a cognitive tax, consuming mental bandwidth that would otherwise be available for long-term planning. In one study, participants reminded of financial stress performed significantly worse on reasoning tasks than those in a control group. ______ the effect disappeared when participants were given time to resolve the hypothetical shortfall before testing, suggesting that the impairment is situational rather than dispositional.\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"Consequently,\",\"B\":\"In addition,\",\"C\":\"Notably,\",\"D\":\"For example,\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m2-q29\",\"number\":29,\"type\":\"transitions\",\"passage\":\"Astronomers once assumed that the rotation rates of spiral galaxies would decline sharply with distance from the galactic center, mirroring the Keplerian falloff observed in our solar system. Vera Rubin's measurements of the Andromeda Galaxy in the 1970s revealed the opposite pattern: stars at the galaxy's outer edge orbited nearly as quickly as those near its core. ______ theorists proposed that vast halos of unseen matter must surround visible galaxies, providing the additional gravitational pull needed to explain the flattened rotation curves.\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"Nevertheless,\",\"B\":\"Similarly,\",\"C\":\"To account for this discrepancy,\",\"D\":\"In contrast,\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m2-q30\",\"number\":30,\"type\":\"transitions\",\"passage\":\"Political philosopher Hannah Arendt distinguished sharply between the private realm of the household and the public realm of political action, arguing that genuine freedom could exist only in the latter. Modern democracies, she contended, have blurred this distinction by treating citizens primarily as consumers whose welfare is managed by administrative bureaucracies. ______ Arendt maintained that reinvigorating the public sphere required reviving practices of deliberation and collective judgment that resist reduction to economic calculation.\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"Accordingly,\",\"B\":\"Conversely,\",\"C\":\"Meanwhile,\",\"D\":\"Granted,\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"}]" }, { "costUsd": 0.122865, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 3 SAT Reading & Writing questions.\nType: rhetorical_synthesis | Test: 11 | Unit: 2 | Start Q#: 31\n\nMANDATORY: Your JSON array MUST contain exactly 3 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m2-qN\" where N is the sequential question number starting at 31.\nFor example, the first question in this batch has id \"sat11-rw-m2-q31\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m2-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 3 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m2-q31\",\"number\":31,\"type\":\"rhetorical_synthesis\",\"passage\":\"While researching the acoustics of stringed instruments, a student has taken the following notes:\\n- The Stradivarius violins were crafted by Antonio Stradivari in Cremona, Italy, between 1666 and 1737.\\n- A 2017 double-blind study by Claudia Fritz tested 137 listeners in a concert hall.\\n- Listeners compared six Stradivarius violins with six newly made violins.\\n- Listeners preferred the sound of the new violins over the Stradivarius instruments.\\n- The result challenges the long-held belief that Stradivarius violins possess uniquely superior tonal qualities.\",\"prompt\":\"The student wants to emphasize the surprising outcome of the 2017 study. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"Antonio Stradivari crafted violins in Cremona, Italy, between 1666 and 1737, and his instruments have long been considered superior.\",\"B\":\"Although Stradivarius violins have long been prized for their supposedly unmatched tone, listeners in Claudia Fritz's 2017 double-blind study actually preferred the sound of newly made violins.\",\"C\":\"In a 2017 study conducted by Claudia Fritz, 137 listeners in a concert hall compared six Stradivarius violins with six newly made violins.\",\"D\":\"Stradivarius violins, crafted in Cremona between 1666 and 1737, were compared with newly made violins in a study involving 137 listeners.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q32\",\"number\":32,\"type\":\"rhetorical_synthesis\",\"passage\":\"While researching bioluminescent organisms, a student has taken the following notes:\\n- Dinoflagellates are single-celled marine plankton.\\n- Some species, such as Noctiluca scintillans, emit blue-green light when disturbed.\\n- The light is produced through a chemical reaction involving luciferin and the enzyme luciferase.\\n- Researchers hypothesize that this flash startles predators, giving the dinoflagellate time to escape.\\n- The phenomenon is often called the 'burglar alarm' defense.\",\"prompt\":\"The student wants to explain the proposed function of bioluminescence in Noctiluca scintillans. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"Noctiluca scintillans is a single-celled marine plankton that produces a blue-green light through a chemical reaction.\",\"B\":\"The light produced by Noctiluca scintillans is generated by a chemical reaction involving luciferin and luciferase.\",\"C\":\"When disturbed, Noctiluca scintillans emits a flash of light that researchers believe startles predators, allowing the dinoflagellate to escape—a mechanism known as the 'burglar alarm' defense.\",\"D\":\"Bioluminescent dinoflagellates, including Noctiluca scintillans, are single-celled organisms that emit blue-green light when disturbed by passing predators.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m2-q33\",\"number\":33,\"type\":\"rhetorical_synthesis\",\"passage\":\"While researching the Harlem Renaissance, a student has taken the following notes:\\n- Zora Neale Hurston (1891-1960) was an American author and anthropologist.\\n- Her novel Their Eyes Were Watching God was published in 1937.\\n- The novel received mixed reviews upon its release.\\n- Richard Wright criticized it for lacking a clear political agenda.\\n- The novel was largely forgotten until Alice Walker rediscovered it in the 1970s.\\n- It is now considered a foundational work of African American literature.\",\"prompt\":\"The student wants to emphasize a contrast between the novel's initial reception and its current status. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"Zora Neale Hurston's Their Eyes Were Watching God, published in 1937, received mixed reviews upon its release, though Richard Wright criticized it for lacking a clear political agenda.\",\"B\":\"Their Eyes Were Watching God, written by Zora Neale Hurston and published in 1937, was rediscovered by Alice Walker in the 1970s.\",\"C\":\"Although Alice Walker rediscovered Their Eyes Were Watching God in the 1970s, the novel had initially been criticized by Richard Wright.\",\"D\":\"Though initially met with mixed reviews and largely forgotten after its 1937 publication, Zora Neale Hurston's Their Eyes Were Watching God is now regarded as a foundational work of African American literature.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"D\"}]" } ], "results": [ { "attempts": 1, "index": 0, "output": { "costUsd": 0.084615, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Reading & Writing questions.\nType: vocabulary_in_context | Test: 11 | Unit: 1 | Start Q#: 1\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m1-qN\" where N is the sequential question number starting at 1.\nFor example, the first question in this batch has id \"sat11-rw-m1-q1\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m1-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m1-q1\",\"number\":1,\"type\":\"vocabulary_in_context\",\"passage\":\"In the decades following the publication of Keynes's General Theory, economists were reluctant to ______ the orthodoxy of classical market equilibrium outright. Instead, they tended to graft Keynesian insights onto existing models, preserving much of the older framework even as its foundational assumptions were being quietly eroded.\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"repudiate\",\"B\":\"disregard\",\"C\":\"misconstrue\",\"D\":\"lament\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m1-q2\",\"number\":2,\"type\":\"vocabulary_in_context\",\"passage\":\"Although Giorgio Vasari praised Pontormo's late frescoes for their technical daring, later critics found the figures' elongated limbs and unnatural palette deeply ______. To these observers, the Mannerist departure from High Renaissance proportion signaled not innovation but a troubling loss of classical composure.\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"prosaic\",\"B\":\"derivative\",\"C\":\"disconcerting\",\"D\":\"ostentatious\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m1-q3\",\"number\":3,\"type\":\"vocabulary_in_context\",\"passage\":\"Behavioral economists have long observed that consumers do not weigh gains and losses symmetrically; a loss of \\\\$100 tends to ______ the pleasure of a \\\\$100 gain. This asymmetry, known as loss aversion, helps explain why investors often hold losing stocks far longer than rational models predict.\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"eclipse\",\"B\":\"replicate\",\"C\":\"supplement\",\"D\":\"anticipate\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m1-q4\",\"number\":4,\"type\":\"vocabulary_in_context\",\"passage\":\"Because the faint signal from the distant quasar was nearly ______ by the glow of intervening galaxies, astronomers at the Keck Observatory employed adaptive optics to isolate its spectrum. Only after filtering out the foreground light could they confirm the object's redshift with any confidence.\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"amplified\",\"B\":\"refracted\",\"C\":\"subsumed\",\"D\":\"occluded\"},\"answer_format\":\"mcq\",\"correct_answer\":\"D\"},{\"id\":\"sat11-rw-m1-q5\",\"number\":5,\"type\":\"vocabulary_in_context\",\"passage\":\"Hannah Arendt argued that totalitarian regimes do not merely suppress dissent; they seek to ______ the very categories through which citizens distinguish truth from falsehood. Once this shared epistemic ground collapses, she warned, resistance itself becomes almost unintelligible.\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"codify\",\"B\":\"dissolve\",\"C\":\"arbitrate\",\"D\":\"venerate\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}]" }, "status": "ok" }, { "attempts": 1, "index": 1, "output": { "costUsd": 0.25419, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 8 SAT Reading & Writing questions.\nType: text_comprehension | Test: 11 | Unit: 1 | Start Q#: 6\n\nMANDATORY: Your JSON array MUST contain exactly 8 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m1-qN\" where N is the sequential question number starting at 6.\nFor example, the first question in this batch has id \"sat11-rw-m1-q6\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m1-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 8 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m1-q6\",\"number\":6,\"type\":\"text_comprehension\",\"passage\":\"The following is adapted from Willa Cather's 1918 novel My Ántonia. The narrator recalls his arrival at his grandparents' Nebraska farm as a child.\\n\\nI had the sense of coming home to myself, and of having found out what a little circle man's experience is. For the first time, perhaps, since that night, I was dissolved into something complete and great. When it came to me, it came as naturally as sleep. The wind sang a sort of clear tune through the long grass, and I lay down among the pumpkins and did not try to think of anything.\",\"prompt\":\"Which choice best describes the narrator's state of mind in the passage?\",\"choices\":{\"A\":\"He feels a profound sense of peace that quiets his impulse toward reflection.\",\"B\":\"He is disoriented by the unfamiliar landscape and seeks refuge among the pumpkins.\",\"C\":\"He is eager to explore the farm but exhausted from his long journey.\",\"D\":\"He is nostalgic for a former home he has recently left behind.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m1-q7\",\"number\":7,\"type\":\"text_comprehension\",\"passage\":\"In a 2019 study, ecologist Mariana Altrichter examined the role of white-lipped peccaries in the Gran Chaco forests of Paraguay. Altrichter found that the peccaries, which travel in herds of up to three hundred, create wallows—shallow muddy depressions—that persist long after the herds have moved on. These wallows retain water during dry seasons and become refuges for frogs, insects, and small mammals that would otherwise perish. When peccary populations collapse due to hunting, the wallows dry up within a single season, and local biodiversity declines sharply.\",\"prompt\":\"Which choice best states the main idea of the text?\",\"choices\":{\"A\":\"Peccary herds in the Gran Chaco have grown in size as hunting pressures have decreased in recent decades.\",\"B\":\"Altrichter's 2019 study was the first to document the behavior of white-lipped peccaries in Paraguay.\",\"C\":\"The disappearance of peccaries threatens the frogs and insects that share their habitat in the Gran Chaco.\",\"D\":\"By creating wallows that sustain other species, peccaries function as ecosystem engineers whose loss reverberates through the Chaco food web.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"D\"},{\"id\":\"sat11-rw-m1-q8\",\"number\":8,\"type\":\"text_comprehension\",\"passage\":\"The following is adapted from a 2021 essay on the composer Florence Price. After Price's death in 1953, most of her manuscripts were stored in an abandoned summer home outside St. Anne, Illinois. In 2009, renovators preparing the house for demolition discovered trunks of her scores, including two violin concertos and the Fourth Symphony, works that had never been performed publicly. The recovery has prompted orchestras to reevaluate a composer who, during her lifetime, was celebrated in Chicago but largely ignored by the East Coast musical establishment.\",\"prompt\":\"The author's main purpose in the passage is to\",\"choices\":{\"A\":\"argue that Price's Fourth Symphony deserves a place in the standard orchestral repertoire.\",\"B\":\"describe a chance discovery that has reshaped the reception of a previously overlooked composer.\",\"C\":\"contrast the musical cultures of Chicago and the East Coast in the early twentieth century.\",\"D\":\"criticize the institutions that failed to preserve Price's manuscripts during her lifetime.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m1-q9\",\"number\":9,\"type\":\"text_comprehension\",\"passage\":\"In The Federalist No. 10 (1787), James Madison addressed the problem of \\\"faction\\\"—groups of citizens united by a common interest adverse to the rights of others or to the public good. Rather than attempting to eliminate factions by suppressing liberty or enforcing uniformity of opinion, Madison argued that the remedy lay in controlling their effects. A large republic, he reasoned, would encompass so many competing interests that no single faction could easily assemble a majority, and elected representatives would tend to refine the raw preferences of constituents into more deliberate judgments.\",\"prompt\":\"Based on the text, which choice best describes Madison's view of factions?\",\"choices\":{\"A\":\"Factions are an unavoidable consequence of liberty, and their dangers are best managed through the structural features of a large republic.\",\"B\":\"Factions arise chiefly from economic inequality, which representative government is well equipped to correct.\",\"C\":\"Factions pose little threat so long as elected representatives share the moral values of their constituents.\",\"D\":\"Factions should be suppressed through legal restrictions that preserve uniformity of opinion among citizens.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m1-q10\",\"number\":10,\"type\":\"text_comprehension\",\"passage\":\"The following is adapted from Edith Wharton's 1905 novel The House of Mirth. Lily Bart, a young socialite, has just encountered Lawrence Selden at a train station.\\n\\nSelden had in fact recognized her at once, and he now advanced toward her with a certain hesitancy. It was characteristic of her to express herself so quickly, and of him to measure his words before committing them to speech. He wondered, as he looked at her, how many years it would take her to learn what it had taken him so long to unlearn: that life, at its deepest, rewarded not the swift gesture but the patient observation.\",\"prompt\":\"Which choice best describes the function of the underlined portion (\\\"He wondered...observation\\\") in the text as a whole?\",\"choices\":{\"A\":\"It reveals Selden's irritation at Lily's impulsive manner and his desire to correct her.\",\"B\":\"It suggests that Selden has come to value reflection over spontaneity and senses that Lily has not yet done so.\",\"C\":\"It implies that Selden envies Lily's quickness and regrets his own deliberate temperament.\",\"D\":\"It indicates that Selden and Lily share a temperament that both have cultivated over many years.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m1-q11\",\"number\":11,\"type\":\"text_comprehension\",\"passage\":\"Astrophysicist Wendy Freedman has long maintained that the Hubble constant—a measure of the universe's expansion rate—cannot be pinned down without first resolving systematic errors in the distance ladder used to calibrate it. In a 2023 paper, Freedman and her collaborators measured distances to nearby galaxies using the tip of the red giant branch (TRGB), a technique that bypasses Cepheid variable stars. Their value of 69.8 km/s/Mpc falls between the higher figure derived from Cepheid-based measurements and the lower figure inferred from the cosmic microwave background, a discrepancy known as the \\\"Hubble tension.\\\"\",\"prompt\":\"Based on the text, what can most reasonably be inferred about the \\\"Hubble tension\\\"?\",\"choices\":{\"A\":\"Freedman's 2023 measurement definitively resolves the Hubble tension in favor of cosmic microwave background estimates.\",\"B\":\"The Hubble tension will likely be resolved once astronomers abandon the use of Cepheid variables as distance indicators.\",\"C\":\"Freedman's intermediate value suggests that unaccounted systematic errors, rather than new physics, may underlie the discrepancy.\",\"D\":\"The Hubble tension arises primarily because the cosmic microwave background cannot be observed at sufficient resolution.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m1-q12\",\"number\":12,\"type\":\"text_comprehension\",\"passage\":\"In studies of intertemporal choice, behavioral economists have long observed that people discount future rewards more steeply than classical models predict. Offered $50 today or $55 in a week, many subjects take the $50; offered $50 in a year or $55 in a year and a week, the same subjects wait. This pattern, termed \\\"hyperbolic discounting,\\\" is difficult to reconcile with the assumption that preferences remain stable across time. Richard Thaler and others have argued that the inconsistency reflects not irrationality but a conflict between a patient \\\"planner\\\" self and an impulsive \\\"doer\\\" self, each exerting influence at different moments.\",\"prompt\":\"Which choice best describes the overall structure of the text?\",\"choices\":{\"A\":\"It identifies a puzzling behavioral pattern and then summarizes one theoretical account offered to explain it.\",\"B\":\"It presents a longstanding economic theory and then describes experimental evidence that decisively refutes it.\",\"C\":\"It compares two competing schools of thought within behavioral economics and favors one over the other.\",\"D\":\"It traces the historical development of hyperbolic discounting from its origins to its current applications.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m1-q13\",\"number\":13,\"type\":\"text_comprehension\",\"passage\":\"Art historian Patricia Rubin has argued that the workshop of Andrea del Verrocchio, active in Florence during the 1470s, functioned less as a training ground for imitators than as a laboratory in which young artists—including the young Leonardo da Vinci—tested new techniques against one another. Surviving panels from the workshop often bear the hands of multiple contributors, and Rubin contends that Verrocchio deliberately assigned passages of the same painting to different apprentices so that their divergent solutions could be studied and compared.\",\"prompt\":\"Which choice best describes the author's tone in the passage?\",\"choices\":{\"A\":\"Skeptical of Rubin's interpretation and inclined to offer an alternative.\",\"B\":\"Neutral in summarizing Rubin's argument without endorsing or rejecting it.\",\"C\":\"Admiring of Verrocchio's business acumen in running a profitable workshop.\",\"D\":\"Dismissive of earlier scholarship on Renaissance apprenticeship practices.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}]" }, "status": "ok" }, { "attempts": 1, "index": 2, "output": { "costUsd": 0.146565, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 4 SAT Reading & Writing questions.\nType: data_analysis | Test: 11 | Unit: 1 | Start Q#: 14\n\nMANDATORY: Your JSON array MUST contain exactly 4 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m1-qN\" where N is the sequential question number starting at 14.\nFor example, the first question in this batch has id \"sat11-rw-m1-q14\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m1-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 4 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m1-q14\",\"number\":14,\"type\":\"data_analysis\",\"passage\":\"Marine biologist Hina Watanabe investigated how water temperature affects the foraging efficiency of juvenile cuttlefish (Sepia officinalis). She measured the average number of successful prey captures per hour for cuttlefish housed in tanks set to four different temperatures, recording results across a six-week observation period.\",\"visual\":{\"type\":\"table\",\"title\":\"Average Prey Captures per Hour by Water Temperature\",\"headers\":[\"Water Temperature (°C)\",\"Average Captures per Hour\",\"Standard Deviation\"],\"rows\":[[\"14\",\"3.2\",\"0.6\"],[\"18\",\"5.8\",\"0.7\"],[\"22\",\"7.1\",\"0.5\"],[\"26\",\"4.4\",\"0.9\"]]},\"prompt\":\"Which statement about the relationship between water temperature and foraging efficiency is best supported by the data?\",\"choices\":{\"A\":\"Foraging efficiency increases steadily as water temperature rises across the entire tested range.\",\"B\":\"Foraging efficiency peaks at an intermediate temperature and declines at both lower and higher temperatures.\",\"C\":\"Foraging efficiency is highest at the coldest temperature tested and decreases as the water warms.\",\"D\":\"Foraging efficiency remains essentially constant regardless of changes in water temperature.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m1-q15\",\"number\":15,\"type\":\"data_analysis\",\"passage\":\"In a behavioral economics study, researcher Diego Alarcón asked participants to choose between receiving a smaller monetary reward immediately or a larger reward after a delay. He recorded the percentage of participants who chose the delayed option at each combination of delay length and reward size, hypothesizing that participants would tolerate longer delays only when the delayed reward was substantially larger than the immediate one.\",\"visual\":{\"type\":\"table\",\"title\":\"Percentage Choosing Delayed Reward over $20 Immediate Payment\",\"headers\":[\"Delay Length\",\"Delayed Reward $25\",\"Delayed Reward $50\",\"Delayed Reward $100\"],\"rows\":[[\"1 week\",\"62%\",\"88%\",\"95%\"],[\"1 month\",\"31%\",\"64%\",\"82%\"],[\"6 months\",\"9%\",\"28%\",\"57%\"]]},\"prompt\":\"Which statement about the data is most consistent with Alarcón's hypothesis?\",\"choices\":{\"A\":\"At every delay length, a larger delayed reward produced a higher percentage of participants willing to wait.\",\"B\":\"Participants always preferred the immediate $20 reward over any delayed reward offered in the study.\",\"C\":\"The percentage of participants choosing the delayed reward depended only on the delay length, not on the reward size.\",\"D\":\"Increasing the delayed reward from $25 to $100 had no measurable effect when the delay was 6 months.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m1-q16\",\"number\":16,\"type\":\"data_analysis\",\"passage\":\"Agronomist Felipe Ortega tested how nitrogen fertilizer application affects yield in a variety of winter wheat. He applied four different rates of nitrogen to otherwise identical plots and recorded the average grain yield, measured in bushels per acre, at the end of the growing season. Ortega claimed that beyond a certain rate, additional nitrogen produces only minimal gains in yield.\",\"visual\":{\"type\":\"table\",\"title\":\"Winter Wheat Yield by Nitrogen Application Rate\",\"headers\":[\"Nitrogen Applied (lb/acre)\",\"Average Yield (bushels/acre)\"],\"rows\":[[\"40\",\"52\"],[\"80\",\"71\"],[\"120\",\"83\"],[\"160\",\"85\"]]},\"prompt\":\"Which finding from the table best supports Ortega's claim?\",\"choices\":{\"A\":\"Increasing nitrogen from 40 to 80 lb/acre raised yield by 19 bushels per acre.\",\"B\":\"Increasing nitrogen from 80 to 120 lb/acre raised yield by 12 bushels per acre.\",\"C\":\"Increasing nitrogen from 120 to 160 lb/acre raised yield by only 2 bushels per acre.\",\"D\":\"The plot receiving 40 lb/acre produced the lowest yield in the study.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m1-q17\",\"number\":17,\"type\":\"data_analysis\",\"passage\":\"Sociologist Marta Kowalski surveyed adults in four metropolitan areas about their weekly volunteer hours and their reported sense of community belonging, scored on a 0-to-10 scale. She wanted to determine whether residents who volunteered more also tended to report a stronger connection to their neighborhood, while cautioning that her data could not establish a causal direction.\",\"visual\":{\"type\":\"table\",\"title\":\"Average Volunteer Hours and Belonging Scores by City\",\"headers\":[\"City\",\"Avg. Weekly Volunteer Hours\",\"Avg. Belonging Score (0-10)\"],\"rows\":[[\"Riverton\",\"1.2\",\"4.3\"],[\"Glenford\",\"2.5\",\"5.9\"],[\"Pinehill\",\"3.8\",\"7.2\"],[\"Westmere\",\"5.1\",\"8.4\"]]},\"prompt\":\"Which conclusion is most strongly supported by Kowalski's data, given her stated caution?\",\"choices\":{\"A\":\"Volunteering more hours each week causes residents to feel a stronger sense of community belonging.\",\"B\":\"Cities with higher average volunteer hours tend to have higher average belonging scores, but the data alone cannot show that one variable causes the other.\",\"C\":\"Residents in cities with low belonging scores would increase their belonging if they were required to volunteer more.\",\"D\":\"Belonging scores and volunteer hours appear to be unrelated across the four cities surveyed.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}]" }, "status": "ok" }, { "attempts": 1, "index": 3, "output": { "costUsd": 0.11679, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 4 SAT Reading & Writing questions.\nType: claim_support | Test: 11 | Unit: 1 | Start Q#: 18\n\nMANDATORY: Your JSON array MUST contain exactly 4 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m1-qN\" where N is the sequential question number starting at 18.\nFor example, the first question in this batch has id \"sat11-rw-m1-q18\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m1-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 4 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"Archaeologist Dr. Marisela Ortega has proposed that the inhabitants of the pre-Columbian settlement of Tiwanaku engaged in long-distance trade with coastal communities as early as 500 CE, rather than relying solely on local resources as previously assumed.\",\"prompt\":\"Which finding, if true, would most directly support Dr. Ortega's hypothesis?\",\"choices\":{\"A\":\"Stone tools found at Tiwanaku exhibit craftsmanship techniques similar to those documented at other highland sites of the same period.\",\"B\":\"Ceramic vessels excavated from Tiwanaku contain residues of marine fish and mollusks that are not native to the highland region.\",\"C\":\"Agricultural terraces near Tiwanaku show evidence of sophisticated irrigation systems that could support a large population.\",\"D\":\"Burial sites at coastal communities contain ornamental objects whose stylistic features differ markedly from those at Tiwanaku.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m1-q19\",\"number\":19,\"type\":\"claim_support\",\"passage\":\"In her 2019 study of urban songbird populations, ornithologist Priya Chandran argued that house finches living in cities have adapted their vocalizations to compensate for low-frequency traffic noise, allowing them to communicate more effectively than their rural counterparts in noisy environments.\",\"prompt\":\"Which finding from Chandran's study would most directly support her claim?\",\"choices\":{\"A\":\"Urban house finches produce songs with higher minimum frequencies than rural house finches, and these songs travel farther when ambient traffic noise is present.\",\"B\":\"Rural house finches and urban house finches exhibit nearly identical body sizes, plumage patterns, and foraging behaviors across the study sites.\",\"C\":\"Urban house finches sing more frequently during early morning hours, when traffic volume in the surrounding neighborhoods tends to be at its lowest.\",\"D\":\"Rural house finch populations have remained stable over the past decade, while urban populations have fluctuated in response to habitat changes.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m1-q20\",\"number\":20,\"type\":\"claim_support\",\"passage\":\"A literary scholar contends that in Virginia Woolf's novel Mrs. Dalloway, the recurring sound of Big Ben functions not merely as a marker of time but as a mechanism that forcibly returns characters from their private reveries to the shared public world of London.\",\"prompt\":\"Which quotation from Mrs. Dalloway would most effectively illustrate the scholar's claim?\",\"choices\":{\"A\":\"\\\"For having lived in Westminster—how many years now? over twenty,—one feels even in the midst of the traffic, or waking at night, Clarissa was positive, a particular hush.\\\"\",\"B\":\"\\\"She had a perpetual sense, as she watched the taxi cabs, of being out, out, far out to sea and alone; she always had the feeling that it was very, very dangerous to live even one day.\\\"\",\"C\":\"\\\"The leaden circles dissolved in the air. Such fools we are, she thought, crossing Victoria Street, and immediately she was back in the noise of the present, hearing the omnibuses grind.\\\"\",\"D\":\"\\\"He had left a note for her, written in pencil, her favourite, saying he would be back; but the words had slipped from her mind like water through fingers.\\\"\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m1-q21\",\"number\":21,\"type\":\"claim_support\",\"passage\":\"Behavioral economist Tomás Reyes hypothesizes that consumers are more likely to complete an online purchase when the checkout process explicitly displays the number of items remaining in stock, because perceived scarcity triggers a sense of urgency that overrides hesitation about the purchase.\",\"prompt\":\"Which finding, if true, would most directly undermine Reyes's hypothesis?\",\"choices\":{\"A\":\"Shoppers presented with \\\"only 3 left in stock\\\" messaging report feeling more anxious during checkout than those who see no such messaging.\",\"B\":\"Websites displaying stock counts generate higher total revenue per visitor than websites that do not display such counts.\",\"C\":\"Completion rates at checkout remain statistically identical whether or not stock-count messaging appears, regardless of how low the count displayed.\",\"D\":\"Consumers who abandon their carts most often cite shipping costs, rather than product availability, as their primary reason for leaving.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"}]" }, "status": "ok" }, { "attempts": 1, "index": 4, "output": { "costUsd": 0.158865, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Reading & Writing questions.\nType: standard_english_conventions | Test: 11 | Unit: 1 | Start Q#: 22\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m1-qN\" where N is the sequential question number starting at 22.\nFor example, the first question in this batch has id \"sat11-rw-m1-q22\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m1-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m1-q22\",\"number\":22,\"type\":\"standard_english_conventions\",\"passage\":\"The extensive collection of botanical illustrations housed at the Royal Horticultural Society's Lindley Library [underlined: are] considered among the finest examples of nineteenth-century scientific artistry in Europe.\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"is\",\"C\":\"were\",\"D\":\"have been\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m1-q23\",\"number\":23,\"type\":\"standard_english_conventions\",\"passage\":\"After reviewing quarterly filings submitted by several multinational firms, the auditor concluded that the [underlined: companies reports] contained irregularities warranting further investigation by regulators.\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"company's reports\",\"C\":\"companies' reports\",\"D\":\"companies reports'\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m1-q24\",\"number\":24,\"type\":\"standard_english_conventions\",\"passage\":\"The nebula NGC 6888, commonly known as the Crescent Nebula, formed approximately 250,000 years [underlined: ago, it was] shaped by powerful stellar winds emanating from a central Wolf-Rayet star.\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"ago, being\",\"C\":\"ago, and\",\"D\":\"ago; it was\"},\"answer_format\":\"mcq\",\"correct_answer\":\"D\"},{\"id\":\"sat11-rw-m1-q25\",\"number\":25,\"type\":\"standard_english_conventions\",\"passage\":\"By the time archaeologist Howard Carter reached the inner chamber of Tutankhamun's tomb in 1923, ancient looters [underlined: had already removed] a small portion of the burial goods, though the bulk of the treasure remained undisturbed.\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"have already removed\",\"C\":\"already remove\",\"D\":\"are already removing\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m1-q26\",\"number\":26,\"type\":\"standard_english_conventions\",\"passage\":\"In her report to the city council, urban planner Yuki Tanaka recommended that the revised zoning policy emphasize transparency, [underlined: encourages] community accountability, and promote equitable access to public resources.\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"encourage\",\"C\":\"encouraging\",\"D\":\"had encouraged\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}]" }, "status": "ok" }, { "attempts": 1, "index": 5, "output": { "costUsd": 0.08049, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 4 SAT Reading & Writing questions.\nType: transitions | Test: 11 | Unit: 1 | Start Q#: 27\n\nMANDATORY: Your JSON array MUST contain exactly 4 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m1-qN\" where N is the sequential question number starting at 27.\nFor example, the first question in this batch has id \"sat11-rw-m1-q27\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m1-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 4 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m1-q27\",\"number\":27,\"type\":\"transitions\",\"passage\":\"The Antikythera mechanism, an ancient Greek device recovered from a shipwreck in 1901, was long assumed to be a simple astronomical clock. ______ recent X-ray tomography has revealed more than thirty interlocking bronze gears capable of predicting eclipses and tracking the irregular motion of the moon, suggesting a level of mechanical sophistication previously thought impossible for the second century BCE.\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"However,\",\"B\":\"Likewise,\",\"C\":\"Consequently,\",\"D\":\"For instance,\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m1-q28\",\"number\":28,\"type\":\"transitions\",\"passage\":\"Economist Elinor Ostrom challenged the prevailing view that shared resources such as fisheries and grazing lands inevitably suffer from overuse. Through decades of fieldwork in Nepal, Kenya, and the Swiss Alps, she documented communities that had sustainably managed common pools for centuries. ______ she identified eight design principles, including clearly defined boundaries and graduated sanctions, that reliably predict which communal arrangements endure.\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"Nevertheless,\",\"B\":\"From these case studies,\",\"C\":\"In contrast,\",\"D\":\"Conversely,\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m1-q29\",\"number\":29,\"type\":\"transitions\",\"passage\":\"The painter Artemisia Gentileschi worked in a profession almost entirely closed to women in seventeenth-century Italy, and her early biographers often framed her career as a curiosity rather than an artistic achievement. Art historian Mary Garrard, writing in the 1980s, argued that this framing obscured the formal rigor of Gentileschi's compositions. ______ her influential monograph reoriented scholarly attention toward the technical choices visible in works such as Judith Slaying Holofernes.\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"Nonetheless,\",\"B\":\"Similarly,\",\"C\":\"Accordingly,\",\"D\":\"In particular,\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m1-q30\",\"number\":30,\"type\":\"transitions\",\"passage\":\"Cuttlefish possess chromatophores, pigment-filled sacs in their skin that expand and contract to produce rapid color changes. These displays are used for camouflage, mating signals, and territorial warnings. ______ cuttlefish are colorblind: their single type of photoreceptor cannot distinguish wavelengths, meaning they coordinate elaborate visual performances they themselves may not fully perceive.\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"Moreover,\",\"B\":\"Therefore,\",\"C\":\"Paradoxically,\",\"D\":\"Meanwhile,\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"}]" }, "status": "ok" }, { "attempts": 1, "index": 6, "output": { "costUsd": 0.11544, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 3 SAT Reading & Writing questions.\nType: rhetorical_synthesis | Test: 11 | Unit: 1 | Start Q#: 31\n\nMANDATORY: Your JSON array MUST contain exactly 3 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m1-qN\" where N is the sequential question number starting at 31.\nFor example, the first question in this batch has id \"sat11-rw-m1-q31\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m1-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 3 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m1-q31\",\"number\":31,\"type\":\"rhetorical_synthesis\",\"passage\":\"While researching the lyrebird, a student has taken these notes:\\n- The superb lyrebird (Menura novaehollandiae) is native to southeastern Australia.\\n- Male lyrebirds are renowned for their vocal mimicry.\\n- A 2021 study by Anastasia Dalziell found that male lyrebirds mimic the alarm calls of multiple bird species during courtship.\\n- Dalziell concluded that this mimicry may deceive females into remaining nearby, mistaking the sounds for a genuine predator threat.\",\"prompt\":\"The student wants to present the conclusion of Dalziell's 2021 study. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"Native to southeastern Australia, the superb lyrebird (Menura novaehollandiae) is famous for the vocal mimicry of its males.\",\"B\":\"In a 2021 study, Anastasia Dalziell observed that male superb lyrebirds mimic the alarm calls of multiple bird species during courtship.\",\"C\":\"Dalziell concluded that male lyrebirds' courtship mimicry may deceive females into staying nearby by simulating a predator threat.\",\"D\":\"Male superb lyrebirds, native to southeastern Australia, are capable of producing the alarm calls of several other bird species.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m1-q32\",\"number\":32,\"type\":\"rhetorical_synthesis\",\"passage\":\"While researching microplastics in marine ecosystems, a student has taken these notes:\\n- Microplastics are plastic fragments smaller than 5 millimeters.\\n- A 2022 survey of the North Atlantic found an average of 1.2 microplastic particles per cubic meter of seawater.\\n- A 2023 survey of the Mediterranean Sea found an average of 4.7 microplastic particles per cubic meter of seawater.\\n- Both surveys used identical sampling methods.\",\"prompt\":\"The student wants to emphasize a difference between the two surveys' findings. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"Microplastics, defined as plastic fragments smaller than 5 millimeters, are now found throughout the world's oceans, including the North Atlantic and the Mediterranean.\",\"B\":\"Although both surveys used identical sampling methods, the 2023 Mediterranean survey detected 4.7 microplastic particles per cubic meter, nearly four times the 1.2 particles per cubic meter recorded in the 2022 North Atlantic survey.\",\"C\":\"Two recent surveys, conducted in 2022 and 2023, measured the concentration of microplastic particles in samples drawn from the North Atlantic and the Mediterranean.\",\"D\":\"Microplastic particles—plastic fragments smaller than 5 millimeters—were detected in both the North Atlantic in 2022 and the Mediterranean in 2023.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m1-q33\",\"number\":33,\"type\":\"rhetorical_synthesis\",\"passage\":\"While researching the painter Artemisia Gentileschi, a student has taken these notes:\\n- Artemisia Gentileschi (1593–1656) was an Italian Baroque painter.\\n- She trained in the workshop of her father, Orazio Gentileschi.\\n- In 1616, she became the first woman admitted to the Accademia delle Arti del Disegno in Florence.\\n- Her paintings frequently depicted heroic women from biblical and mythological narratives.\",\"prompt\":\"The student wants to introduce Artemisia Gentileschi to an audience unfamiliar with her. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"In 1616, she became the first woman admitted to the Accademia delle Arti del Disegno in Florence.\",\"B\":\"Trained in her father Orazio's workshop, she frequently depicted heroic women from biblical and mythological narratives.\",\"C\":\"Artemisia Gentileschi (1593–1656) was an Italian Baroque painter known for her depictions of heroic women from biblical and mythological narratives.\",\"D\":\"Artemisia Gentileschi's paintings, which often featured biblical and mythological heroines, reflected the techniques she learned in her father's workshop.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"}]" }, "status": "ok" }, { "attempts": 1, "index": 7, "output": { "costUsd": 0.08679, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Reading & Writing questions.\nType: vocabulary_in_context | Test: 11 | Unit: 2 | Start Q#: 1\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m2-qN\" where N is the sequential question number starting at 1.\nFor example, the first question in this batch has id \"sat11-rw-m2-q1\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m2-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m2-q1\",\"number\":1,\"type\":\"vocabulary_in_context\",\"passage\":\"In behavioral economics, researchers often find that small contextual cues can ______ decisions that individuals believe are made on purely rational grounds. A series of experiments by Thaler and Sunstein demonstrated that default options on retirement forms, though seemingly trivial, shifted enrollment rates by more than forty percent.\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"dictate\",\"B\":\"inflect\",\"C\":\"fabricate\",\"D\":\"commandeer\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q2\",\"number\":2,\"type\":\"vocabulary_in_context\",\"passage\":\"Although the fresco cycle in the Brancacci Chapel is now attributed largely to Masaccio, earlier attributions were often ______: nineteenth-century connoisseurs assigned individual panels to Masolino or Filippino Lippi on the basis of stylistic hunches that modern pigment analysis has since overturned.\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"conjectural\",\"B\":\"duplicitous\",\"C\":\"indifferent\",\"D\":\"exhaustive\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m2-q3\",\"number\":3,\"type\":\"vocabulary_in_context\",\"passage\":\"The protein chaperone Hsp90 does not merely fold its client molecules; it also ______ their activity by holding them in metastable conformations until a cellular signal permits full maturation. This capacity to pause and then release function is central to how eukaryotic cells cope with environmental stress.\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"abrogates\",\"B\":\"amplifies\",\"C\":\"modulates\",\"D\":\"misconstrues\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m2-q4\",\"number\":4,\"type\":\"vocabulary_in_context\",\"passage\":\"In $\\\\textit{The Social Contract}$, Rousseau argues that legitimate political authority cannot be ______ from any natural hierarchy; rather, it must arise from the free consent of equals who agree to be bound by a general will. This claim marked a decisive break from earlier theorists who grounded sovereignty in paternal or divine precedence.\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"rescinded\",\"B\":\"absolved\",\"C\":\"withheld\",\"D\":\"derived\"},\"answer_format\":\"mcq\",\"correct_answer\":\"D\"},{\"id\":\"sat11-rw-m2-q5\",\"number\":5,\"type\":\"vocabulary_in_context\",\"passage\":\"Observations from the James Webb Space Telescope have begun to ______ long-standing models of galaxy formation: massive, well-structured galaxies appear at redshifts so high that, under prevailing theories, they should not yet have had time to assemble. Cosmologists are now revisiting assumptions about the efficiency of early star formation.\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"unsettle\",\"B\":\"annihilate\",\"C\":\"ratify\",\"D\":\"postpone\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"}]" }, "status": "ok" }, { "attempts": 1, "index": 8, "output": { "costUsd": 0.24009, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 8 SAT Reading & Writing questions.\nType: text_comprehension | Test: 11 | Unit: 2 | Start Q#: 6\n\nMANDATORY: Your JSON array MUST contain exactly 8 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m2-qN\" where N is the sequential question number starting at 6.\nFor example, the first question in this batch has id \"sat11-rw-m2-q6\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m2-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 8 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m2-q6\",\"number\":6,\"type\":\"text_comprehension\",\"passage\":\"In her 1862 letters, Emily Dickinson confided to her mentor Thomas Wentworth Higginson that she avoided publication because she feared the 'auction' of the mind. Yet she circulated dozens of poems privately, sewing them into small hand-bound booklets she called fascicles. Scholars have argued that these fascicles constitute a form of publication in themselves, one that preserved authorial control while still inviting a select readership.\",\"prompt\":\"Which choice best states the main idea of the text?\",\"choices\":{\"A\":\"Dickinson's refusal to publish her poems reflected a broader Victorian discomfort with female authorship.\",\"B\":\"Dickinson's fascicles can be understood as an alternative mode of publication rather than a retreat from readership.\",\"C\":\"Dickinson's correspondence with Higginson was the primary vehicle through which her poetry reached the public.\",\"D\":\"Dickinson's hand-bound booklets were technically inferior to commercial print but artistically superior.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q7\",\"number\":7,\"type\":\"text_comprehension\",\"passage\":\"The following is adapted from a 2019 article in a marine biology journal. Sea otters (Enhydra lutris) are often described as a keystone species because of their outsized effect on kelp-forest ecosystems. By preying on sea urchins, otters prevent the urchins from overgrazing kelp, which in turn shelters dozens of fish and invertebrate species. Recent surveys along the central California coast, however, suggest that the relationship is more conditional than once assumed: in bays where pycnopodia sea stars—another urchin predator—have collapsed, otters alone cannot suppress urchin populations enough to protect kelp.\",\"prompt\":\"What is the main purpose of the text?\",\"choices\":{\"A\":\"To argue that sea otters have been incorrectly classified as a keystone species in California kelp forests.\",\"B\":\"To describe a predator-prey relationship that has remained stable despite recent ecological disturbances.\",\"C\":\"To qualify a widely held claim about sea otters by presenting evidence that their ecological role depends on context.\",\"D\":\"To compare the hunting behaviors of sea otters and pycnopodia sea stars in protecting kelp forests.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m2-q8\",\"number\":8,\"type\":\"text_comprehension\",\"passage\":\"The following is adapted from Willa Cather's 1918 novel My Ántonia. The narrator Jim Burden recalls his grandmother's garden. I sat down in the middle of the garden, where snakes could scarcely approach unseen, and leaned my back against a warm yellow pumpkin. There were some ground-cherry bushes growing along the furrows, full of fruit. I turned back the papery triangular sheaths that protected the berries and ate a few. All about me giant grasshoppers, twice as big as any I had ever seen, were doing acrobatic feats among the dried vines. The gophers scurried up and down the ploughed ground.\",\"prompt\":\"Which choice best describes the function of the underlined details (the grasshoppers, gophers, and pumpkin) in the passage?\",\"choices\":{\"A\":\"They suggest that the garden has fallen into a state of neglect and disorder.\",\"B\":\"They establish Jim's detailed, sensory immersion in a landscape that feels abundant and alive.\",\"C\":\"They foreshadow a later conflict between Jim and the natural world of the prairie.\",\"D\":\"They contrast Jim's childhood memories with the harsher realities of farm labor.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q9\",\"number\":9,\"type\":\"text_comprehension\",\"passage\":\"Economist Elinor Ostrom, awarded the Nobel Prize in 2009, challenged the then-dominant assumption that shared resources—fisheries, grazing lands, irrigation systems—were inevitably overexploited unless governed by either private ownership or centralized state control. Drawing on fieldwork in villages from Nepal to Maine, she documented communities that had sustainably managed common resources for centuries through locally devised rules and mutual monitoring. Her work did not dispute that commons can collapse; rather, it identified the institutional conditions under which they do not.\",\"prompt\":\"Based on the text, how did Ostrom's research most likely reshape scholarly thinking about common resources?\",\"choices\":{\"A\":\"It demonstrated that privatization is the only reliable safeguard against resource depletion.\",\"B\":\"It showed that state regulation and private ownership are equally effective across all contexts.\",\"C\":\"It revealed a third pathway to sustainability that had been largely overlooked by earlier theorists.\",\"D\":\"It established that resource collapse is primarily caused by a failure of centralized enforcement.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m2-q10\",\"number\":10,\"type\":\"text_comprehension\",\"passage\":\"The following is adapted from an essay on Baroque painting. Caravaggio's 'The Calling of Saint Matthew' (1599–1600) resolves its dramatic tension through a single diagonal shaft of light that cuts across the shadowed tavern. Where earlier Mannerist painters crowded the picture plane with writhing figures and competing gestures, Caravaggio does the opposite: he empties the scene of ornament so that the viewer's eye has nowhere to travel but along the beam of light. What might in lesser hands have become theatrical instead becomes inevitable.\",\"prompt\":\"As used in the final sentence, 'inevitable' most nearly suggests that the composition feels\",\"choices\":{\"A\":\"oppressive and overburdened by symbolic weight.\",\"B\":\"compositionally necessary rather than merely staged.\",\"C\":\"derivative of earlier Mannerist conventions.\",\"D\":\"ambiguous in its spiritual implications.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q11\",\"number\":11,\"type\":\"text_comprehension\",\"passage\":\"A 2021 study of urban foxes in London by ecologist Kevin Parnell reported that foxes living within the M25 orbital motorway had, over roughly seventy generations, developed shorter, broader snouts and smaller brain cases than their rural counterparts. Parnell argued that these changes parallel the cranial shifts associated with domestication in dogs and cats, even though no human has deliberately bred the foxes. He suggested that the sustained proximity to humans, and the reduced selective pressure for wariness, may be producing a form of self-domestication.\",\"prompt\":\"Which choice best describes the structure of the text?\",\"choices\":{\"A\":\"A long-standing scientific theory is presented and then rejected in light of new genetic evidence.\",\"B\":\"A researcher's findings are summarized, followed by the interpretive framework he proposes to explain them.\",\"C\":\"Two competing explanations for an observed phenomenon are compared and one is endorsed.\",\"D\":\"An anecdotal observation is generalized into a claim about urban wildlife across Europe.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q12\",\"number\":12,\"type\":\"text_comprehension\",\"passage\":\"Historian Natalie Zemon Davis, writing about sixteenth-century France, once remarked that the archival record is 'not a mirror but a sieve.' What reaches the historian, she argued, has already been filtered by the interests of those who kept records: notaries, tax collectors, church officials. A peasant woman's dispute over a dowry survives only because a court clerk found it worth transcribing. The historian's task, then, is less to reconstruct a vanished totality than to read attentively against the grain of the sources that happen to remain.\",\"prompt\":\"Which choice best states the central claim of the text?\",\"choices\":{\"A\":\"Sixteenth-century French archives are too incomplete to support reliable historical scholarship.\",\"B\":\"Historians should prioritize the recovery of voices belonging to peasant women over those of clerks.\",\"C\":\"Because archival sources reflect the priorities of their creators, historians must interpret them critically rather than treating them as transparent windows onto the past.\",\"D\":\"The metaphor of the sieve is more accurate than the metaphor of the mirror in describing any form of historical evidence.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m2-q13\",\"number\":13,\"type\":\"text_comprehension\",\"passage\":\"In studies of infant cognition, researchers often rely on the 'violation of expectation' paradigm: infants are shown an event that either conforms to or breaks a physical rule, such as an object passing through a solid wall. Infants who look longer at the impossible event are presumed to have expected the possible one. Critics have noted, however, that longer looking times could reflect any number of reactions—surprise, confusion, visual novelty—and that inferring specific conceptual knowledge from gaze alone may credit infants with more than the data warrant.\",\"prompt\":\"Which choice best describes the author's attitude toward the violation-of-expectation paradigm?\",\"choices\":{\"A\":\"Enthusiastic endorsement of its power to reveal innate knowledge.\",\"B\":\"Cautious skepticism about the specificity of the inferences it supports.\",\"C\":\"Dismissive rejection of its value to developmental research.\",\"D\":\"Neutral description without evaluative commentary.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}]" }, "status": "ok" }, { "attempts": 1, "index": 9, "output": { "costUsd": 0.344415, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 4 SAT Reading & Writing questions.\nType: data_analysis | Test: 11 | Unit: 2 | Start Q#: 14\n\nMANDATORY: Your JSON array MUST contain exactly 4 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m2-qN\" where N is the sequential question number starting at 14.\nFor example, the first question in this batch has id \"sat11-rw-m2-q14\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m2-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 4 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m2-q14\",\"number\":14,\"type\":\"data_analysis\",\"passage\":\"A team of plant physiologists measured the weekly growth of Boston fern specimens (Nephrolepis exaltata) cultivated under five different light intensities, holding temperature, humidity, and nutrients constant. A student summarizing the experiment writes: \\\"Although the fern responds strongly to modest increases in light at low intensities, ______\\\"\",\"visual\":{\"type\":\"table\",\"title\":\"Boston Fern Growth Rate by Light Intensity\",\"headers\":[\"Light Intensity (μmol/m²/s)\",\"Growth Rate (mm/week)\"],\"rows\":[[\"100\",\"4.2\"],[\"200\",\"7.8\"],[\"400\",\"11.5\"],[\"800\",\"12.9\"],[\"1600\",\"13.1\"]]},\"prompt\":\"Which choice most effectively uses data from the table to complete the statement?\",\"choices\":{\"A\":\"growth continues to rise in direct proportion to light intensity as it doubles from 400 to 1600 $\\\\mu mol/m^2/s$.\",\"B\":\"further increases in intensity beyond 400 $\\\\mu mol/m^2/s$ yield progressively smaller gains in growth rate.\",\"C\":\"growth rate declines sharply once light intensity exceeds 800 $\\\\mu mol/m^2/s$.\",\"D\":\"the relationship between light intensity and weekly growth is not evident in the measurements.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q15\",\"number\":15,\"type\":\"data_analysis\",\"passage\":\"In a behavioral-economics field study conducted at a Dutch supermarket, researchers varied the number of jam flavors displayed on a tasting booth and recorded both the percentage of passing shoppers who made a purchase and the percentage of purchasers who later reported being satisfied with their selection. A student summarizing the results claims that presenting more choices consistently reduces both the likelihood that a shopper will buy and the shopper's satisfaction with what was chosen.\",\"visual\":{\"type\":\"table\",\"title\":\"Purchase and Satisfaction Rates by Number of Jam Options\",\"headers\":[\"Number of Options\",\"Shoppers Who Purchased (%)\",\"Purchasers Satisfied with Choice (%)\"],\"rows\":[[\"6\",\"30\",\"84\"],[\"12\",\"22\",\"72\"],[\"18\",\"15\",\"64\"],[\"24\",\"10\",\"55\"]]},\"prompt\":\"Which choice most effectively uses data from the table to support the student's claim?\",\"choices\":{\"A\":\"At 6 options, 84% of purchasers reported satisfaction, while at 24 options only 10% of shoppers made a purchase.\",\"B\":\"As the number of options rose from 6 to 24, both the purchase rate and the satisfaction rate increased steadily.\",\"C\":\"Satisfaction rates among purchasers remained effectively constant regardless of the number of options displayed.\",\"D\":\"As the number of options rose from 6 to 24, the purchase rate fell from 30% to 10% and the satisfaction rate fell from 84% to 55%.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"D\"},{\"id\":\"sat11-rw-m2-q16\",\"number\":16,\"type\":\"data_analysis\",\"passage\":\"Astronomers detect exoplanets primarily through two techniques: the radial velocity (RV) method, which measures the stellar wobble induced by an orbiting planet, and transit photometry (TP), which measures periodic dimming as a planet crosses its star. A recent review tabulated confirmed detections by each method across four planetary mass bins. An astronomy student writes: \\\"Although the radial velocity technique is particularly effective for identifying massive planets, transit photometry has proven far more productive for detecting smaller worlds.\\\"\",\"visual\":{\"type\":\"table\",\"title\":\"Confirmed Exoplanet Detections by Method and Mass\",\"headers\":[\"Planet Mass (Earth masses)\",\"RV Method\",\"TP Method\"],\"rows\":[[\"< 1\",\"12\",\"47\"],[\"1–10\",\"89\",\"203\"],[\"10–100\",\"156\",\"118\"],[\"> 100\",\"74\",\"39\"]]},\"prompt\":\"Which choice most effectively uses data from the table to support the student's claim?\",\"choices\":{\"A\":\"For planets below 10 Earth masses, transit photometry confirmed more than twice as many detections as radial velocity (250 vs. 101), whereas for planets above 100 Earth masses, radial velocity confirmed nearly twice as many as transit photometry (74 vs. 39).\",\"B\":\"Radial velocity detected 12 planets with masses below 1 Earth mass, while transit photometry detected 39 planets with masses above 100 Earth masses.\",\"C\":\"Transit photometry produced more confirmed detections than radial velocity in every mass range listed in the table.\",\"D\":\"The radial velocity method has failed to confirm any planets with masses below 1 Earth mass.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m2-q17\",\"number\":17,\"type\":\"data_analysis\",\"passage\":\"An economist at the OECD compared the tertiary-education wage premium—the percentage by which workers with a university degree out-earn otherwise comparable workers whose schooling ended at the secondary level—for two age cohorts (25–34 and 55–64) in four member countries. A student summarizing the findings claims that in every country surveyed the wage premium conferred by a tertiary degree grows larger over the course of a worker's career.\",\"visual\":{\"type\":\"table\",\"title\":\"Tertiary-Education Wage Premium by Country and Age Cohort (%)\",\"headers\":[\"Country\",\"Ages 25–34\",\"Ages 55–64\"],\"rows\":[[\"Germany\",\"38\",\"76\"],[\"South Korea\",\"32\",\"58\"],[\"United States\",\"57\",\"92\"],[\"Italy\",\"28\",\"49\"]]},\"prompt\":\"Which choice most effectively uses data from the table to support the student's claim?\",\"choices\":{\"A\":\"The United States displayed the highest wage premium among workers aged 55–64, at 92 percent.\",\"B\":\"The wage premium in Italy was smaller than that in Germany for both age cohorts.\",\"C\":\"In each of the four countries, the wage premium for workers aged 55–64 was substantially higher than the premium for workers aged 25–34.\",\"D\":\"South Korea showed a smaller premium than Germany for younger workers but a comparable premium for older workers.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"}]" }, "status": "ok" }, { "attempts": 1, "index": 10, "output": { "costUsd": 0.13599, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 4 SAT Reading & Writing questions.\nType: claim_support | Test: 11 | Unit: 2 | Start Q#: 18\n\nMANDATORY: Your JSON array MUST contain exactly 4 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m2-qN\" where N is the sequential question number starting at 18.\nFor example, the first question in this batch has id \"sat11-rw-m2-q18\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m2-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 4 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m2-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A team of marine biologists studying the Pacific bluefin tuna (Thunnus orynnalis) hypothesized that the species' ability to maintain a body temperature significantly warmer than the surrounding seawater allows it to forage efficiently in colder, deeper waters where prey density is higher. To test this, the team tagged 84 individuals with biologgers that recorded muscle temperature, ambient water temperature, and depth over a twelve-month period.\",\"prompt\":\"Which finding from the study, if true, would most directly support the researchers' hypothesis?\",\"choices\":{\"A\":\"Tagged tuna spent approximately equal proportions of time in surface waters and in deeper waters, regardless of the thermal gradient between their muscle tissue and the surrounding environment.\",\"B\":\"Individuals whose muscle temperatures exceeded ambient water temperatures by the largest margins executed significantly more feeding dives into cold, prey-rich waters below 400 meters than did individuals with smaller thermal differentials.\",\"C\":\"The average muscle temperature of tagged tuna was roughly 5 degrees Celsius higher than the ambient water temperature across all depths sampled during the study period.\",\"D\":\"Tuna that remained primarily in warm surface waters exhibited higher overall growth rates than those that frequently descended into colder, deeper zones where prey were more abundant.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q19\",\"number\":19,\"type\":\"claim_support\",\"passage\":\"In her 2019 study of Florentine workshop practice, art historian Marta Ricci argues that Andrea del Verrocchio's studio in the 1470s operated less as a conventional master-apprentice hierarchy and more as a collaborative environment in which senior apprentices were entrusted with substantive portions of commissioned works, including passages that bore the master's signature.\",\"prompt\":\"Which finding, if true, would most directly support Ricci's argument?\",\"choices\":{\"A\":\"Surviving account books from Verrocchio's workshop show that the master personally received payment for all commissions completed between 1470 and 1478.\",\"B\":\"Technical analysis of Verrocchio's Baptism of Christ reveals that the angel on the left and portions of the landscape were executed by the young Leonardo da Vinci using a distinctly different oil-glaze technique.\",\"C\":\"Contemporary Florentine guild regulations required that any panel painting bearing a master's signature be certified as having passed inspection by the workshop's head artisan.\",\"D\":\"Verrocchio is documented to have taken on an unusually large number of apprentices during the 1470s, which scholars have attributed to growing demand for sculptural commissions in the city.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q20\",\"number\":20,\"type\":\"claim_support\",\"passage\":\"In his essay collection The Weight of Ordinary Things (2021), critic Daniel Oyelaran contends that the short stories of Yiyun Li resist conventional emotional payoffs: rather than building toward catharsis, her narratives accumulate small, unresolved observations that leave the reader in a state of productive unease.\",\"prompt\":\"Which quotation from a Yiyun Li story would most effectively illustrate Oyelaran's claim?\",\"choices\":{\"A\":\"\\\"When the letter finally arrived, Mei wept openly in the kitchen, and for the first time in years she felt something loosen in her chest, as though a door she had long kept shut had swung wide.\\\"\",\"B\":\"\\\"He noticed the faint stain on the tablecloth, the way his sister's hand hovered near her teacup without lifting it, the silence that was not quite a silence; he said nothing, and neither did she, and eventually the waiter came.\\\"\",\"C\":\"\\\"The villain's confession, delivered in the final pages with theatrical flourish, resolved every question the novel had raised and sent readers home satisfied that justice had at last been served.\\\"\",\"D\":\"\\\"Anger rose in him with sudden clarity, and he understood, as if for the first time, exactly what he must do and why; the path ahead shone bright before him.\\\"\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q21\",\"number\":21,\"type\":\"claim_support\",\"passage\":\"An economist studying regional labor markets claims that increases in a state's minimum wage lead to measurable improvements in worker retention at low-wage firms because the higher wage reduces the incentive for employees to search for better-paying positions elsewhere.\",\"prompt\":\"Which finding, if true, would most directly support the economist's claim?\",\"choices\":{\"A\":\"Following a state-level minimum wage increase, the average voluntary quit rate among workers at low-wage firms in that state declined by 18 percent relative to neighboring states that did not raise their minimum wage.\",\"B\":\"States that raised their minimum wage experienced a slight decline in the total number of low-wage jobs available during the year following the policy change.\",\"C\":\"Surveys indicate that workers who earn above their state's minimum wage generally report higher levels of job satisfaction than workers who earn at or near the minimum.\",\"D\":\"After the minimum wage increase took effect, consumer prices at low-wage firms in affected states rose modestly, partially offsetting the real wage gains for workers.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"}]" }, "status": "ok" }, { "attempts": 1, "index": 11, "output": { "costUsd": 0.095265, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Reading & Writing questions.\nType: standard_english_conventions | Test: 11 | Unit: 2 | Start Q#: 22\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m2-qN\" where N is the sequential question number starting at 22.\nFor example, the first question in this batch has id \"sat11-rw-m2-q22\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m2-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m2-q22\",\"number\":22,\"type\":\"standard_english_conventions\",\"passage\":\"In 1977, marine biologists aboard the submersible Alvin discovered hydrothermal vents near the Galápagos Rift. The communities of tube worms, clams, and crabs clustered around these vents [underlined: thrives] in an environment once thought incapable of supporting life.\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"thrive\",\"C\":\"has thrived\",\"D\":\"thriving\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q23\",\"number\":23,\"type\":\"standard_english_conventions\",\"passage\":\"The Ghanaian architect David Adjaye is known for designs that engage with cultural memory. His firm's most celebrated project, the Smithsonian National Museum of African American History and [underlined: Culture, opened] in 2016 on the National Mall in Washington, D.C.\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"Culture opened,\",\"C\":\"Culture; opened\",\"D\":\"Culture opened\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m2-q24\",\"number\":24,\"type\":\"standard_english_conventions\",\"passage\":\"Paleontologist Mary Anning, who spent much of her life collecting fossils along the cliffs of Lyme Regis, identified the first complete ichthyosaur skeleton when she was only twelve years old. [underlined: Despite her remarkable contributions to geology, she] was denied membership in the Geological Society of London because of her gender.\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"Despite her remarkable contributions to geology she\",\"C\":\"Despite her remarkable contributions to geology; she\",\"D\":\"Despite her, remarkable contributions to geology, she\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"},{\"id\":\"sat11-rw-m2-q25\",\"number\":25,\"type\":\"standard_english_conventions\",\"passage\":\"Economist Elinor Ostrom challenged the prevailing view that shared resources inevitably collapse under private exploitation. Drawing on fieldwork in Nepal, Kenya, and Switzerland, [underlined: communities that manage fisheries, forests, and irrigation systems successfully were documented by Ostrom].\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"fisheries, forests, and irrigation systems were documented as successfully managed by communities in Ostrom's work\",\"C\":\"Ostrom documented communities that successfully manage fisheries, forests, and irrigation systems\",\"D\":\"there was documentation by Ostrom of communities successfully managing fisheries, forests, and irrigation systems\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m2-q26\",\"number\":26,\"type\":\"standard_english_conventions\",\"passage\":\"The Voyager 1 spacecraft, launched in 1977, crossed into interstellar space in 2012. Although its instruments now return data from more than 15 billion miles away, [underlined: their] plutonium power source is expected to fail sometime after 2025.\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"it's\",\"C\":\"its\",\"D\":\"they're\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"}]" }, "status": "ok" }, { "attempts": 1, "index": 12, "output": { "costUsd": 0.08484, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 4 SAT Reading & Writing questions.\nType: transitions | Test: 11 | Unit: 2 | Start Q#: 27\n\nMANDATORY: Your JSON array MUST contain exactly 4 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m2-qN\" where N is the sequential question number starting at 27.\nFor example, the first question in this batch has id \"sat11-rw-m2-q27\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m2-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 4 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m2-q27\",\"number\":27,\"type\":\"transitions\",\"passage\":\"The rediscovery of Sofonisba Anguissola in the late twentieth century prompted museums to reattribute several sixteenth-century portraits long credited to male contemporaries. Curators initially hesitated, citing a lack of signed works by the artist. ______ subsequent archival research uncovered letters from the Spanish court confirming her authorship of numerous royal portraits, including ones previously ascribed to Alonso Sánchez Coello.\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"Likewise,\",\"B\":\"However,\",\"C\":\"Therefore,\",\"D\":\"For instance,\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q28\",\"number\":28,\"type\":\"transitions\",\"passage\":\"Behavioral economist Sendhil Mullainathan has argued that scarcity imposes a cognitive tax, consuming mental bandwidth that would otherwise be available for long-term planning. In one study, participants reminded of financial stress performed significantly worse on reasoning tasks than those in a control group. ______ the effect disappeared when participants were given time to resolve the hypothetical shortfall before testing, suggesting that the impairment is situational rather than dispositional.\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"Consequently,\",\"B\":\"In addition,\",\"C\":\"Notably,\",\"D\":\"For example,\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m2-q29\",\"number\":29,\"type\":\"transitions\",\"passage\":\"Astronomers once assumed that the rotation rates of spiral galaxies would decline sharply with distance from the galactic center, mirroring the Keplerian falloff observed in our solar system. Vera Rubin's measurements of the Andromeda Galaxy in the 1970s revealed the opposite pattern: stars at the galaxy's outer edge orbited nearly as quickly as those near its core. ______ theorists proposed that vast halos of unseen matter must surround visible galaxies, providing the additional gravitational pull needed to explain the flattened rotation curves.\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"Nevertheless,\",\"B\":\"Similarly,\",\"C\":\"To account for this discrepancy,\",\"D\":\"In contrast,\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m2-q30\",\"number\":30,\"type\":\"transitions\",\"passage\":\"Political philosopher Hannah Arendt distinguished sharply between the private realm of the household and the public realm of political action, arguing that genuine freedom could exist only in the latter. Modern democracies, she contended, have blurred this distinction by treating citizens primarily as consumers whose welfare is managed by administrative bureaucracies. ______ Arendt maintained that reinvigorating the public sphere required reviving practices of deliberation and collective judgment that resist reduction to economic calculation.\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"Accordingly,\",\"B\":\"Conversely,\",\"C\":\"Meanwhile,\",\"D\":\"Granted,\"},\"answer_format\":\"mcq\",\"correct_answer\":\"A\"}]" }, "status": "ok" }, { "attempts": 1, "index": 13, "output": { "costUsd": 0.122865, "model": "opus", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 3 SAT Reading & Writing questions.\nType: rhetorical_synthesis | Test: 11 | Unit: 2 | Start Q#: 31\n\nMANDATORY: Your JSON array MUST contain exactly 3 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-rw-m2-qN\" where N is the sequential question number starting at 31.\nFor example, the first question in this batch has id \"sat11-rw-m2-q31\".\n\nFollow the instructions for your type:\n\n[vocabulary_in_context]\n- 2-3 sentence passage with ______ blank from specific academic domains (molecular biology, Renaissance art, behavioral economics, astrophysics, political philosophy)\n- 4 college-level word choices from the same semantic field\n- Correct word is most PRECISE for context, not most common synonym\n- Schema: {\"id\":\"sat11-rw-m2-qN\",\"number\":N,\"type\":\"vocabulary_in_context\",\"passage\":\"...with ______...\",\"prompt\":\"Which choice completes the text with the most logical and precise word or phrase?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[text_comprehension]\n- 3-8 sentence passage from published-quality source (journal, novel, essay) with specific names/dates/findings\n- Mix subtypes: main idea (2-3), author's purpose (1-2), inference (2-3), tone (0-1), structure (1-2)\n- Topics: natural science, social science, literature, history, arts/humanities (use each)\n- Distractors: surface-level misreading, wrong scope, opposite interpretation\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"text_comprehension\",\"passage\":\"...\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[data_analysis]\n- Passage describing realistic data/study results requiring INTERPRETATION, not lookup\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Data patterns: diminishing returns, optimal values, conditional relationships\n- Distractors: wrong column, wrong trend direction, correlation vs causation confusion\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"data_analysis\",\"passage\":\"...\",\"visual\":{...},\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[claim_support]\n- Mix: \"Which finding supports/undermines the claim?\" AND \"Which quotation illustrates the claim?\"\n- Claims must be specific and testable. Distractors are topically related but don't address the specific mechanism.\n- CRITICAL: \"choices\" contains ONLY the 4 keys A, B, C, D. The \"answer_format\" and \"correct_answer\" keys are SIBLINGS of \"choices\", NEVER nested inside it. Close the \"choices\" object with \"}\" then continue with other top-level keys, then close the question object with another \"}\".\n- Example (study this exact structure):\n{\"id\":\"sat99-rw-m1-q18\",\"number\":18,\"type\":\"claim_support\",\"passage\":\"A researcher hypothesizes that X causes Y.\",\"prompt\":\"Which finding supports the hypothesis?\",\"choices\":{\"A\":\"Control finding\",\"B\":\"Supporting finding\",\"C\":\"Neutral finding\",\"D\":\"Counter finding\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"}\n\n[standard_english_conventions]\n- Passage with [underlined: text here] portion. Student picks best replacement.\n- Cover: subject-verb agreement, pronouns, verb tense, parallel structure, modifiers, commas, semicolons, apostrophes, sentence boundaries\n- \"NO CHANGE\" (choice A) correct ~25% of the time, NOT always\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"standard_english_conventions\",\"passage\":\"...with [underlined: ...]...\",\"prompt\":\"Which choice completes the text so that it conforms to the conventions of Standard English?\",\"choices\":{\"A\":\"NO CHANGE\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[transitions]\n- Passage with ______ blank connecting two ideas. All 4 choices grammatically correct; distinction is logical.\n- Cover: contrast, addition, cause/effect, example, sequence, concession\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"transitions\",\"passage\":\"...______ ...\",\"prompt\":\"Which choice completes the text with the most logical transition?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\n[rhetorical_synthesis]\n- Notes format: \"While researching [topic], a student has taken these notes:\\n- Note 1\\n- Note 2\\n- Note 3\"\n- Specific rhetorical goal stated. All 4 choices are grammatical sentences using the notes; only 1 achieves the goal.\n- Goals: emphasize difference, highlight finding, introduce topic, summarize conclusion, present cause-effect\n- Schema: {\"id\":\"...\",\"number\":N,\"type\":\"rhetorical_synthesis\",\"passage\":\"...\",\"prompt\":\"The student wants to [goal]. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\n\nOutput ONLY a valid JSON array of EXACTLY 3 objects. No markdown. Count them before outputting.", "text": "[{\"id\":\"sat11-rw-m2-q31\",\"number\":31,\"type\":\"rhetorical_synthesis\",\"passage\":\"While researching the acoustics of stringed instruments, a student has taken the following notes:\\n- The Stradivarius violins were crafted by Antonio Stradivari in Cremona, Italy, between 1666 and 1737.\\n- A 2017 double-blind study by Claudia Fritz tested 137 listeners in a concert hall.\\n- Listeners compared six Stradivarius violins with six newly made violins.\\n- Listeners preferred the sound of the new violins over the Stradivarius instruments.\\n- The result challenges the long-held belief that Stradivarius violins possess uniquely superior tonal qualities.\",\"prompt\":\"The student wants to emphasize the surprising outcome of the 2017 study. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"Antonio Stradivari crafted violins in Cremona, Italy, between 1666 and 1737, and his instruments have long been considered superior.\",\"B\":\"Although Stradivarius violins have long been prized for their supposedly unmatched tone, listeners in Claudia Fritz's 2017 double-blind study actually preferred the sound of newly made violins.\",\"C\":\"In a 2017 study conducted by Claudia Fritz, 137 listeners in a concert hall compared six Stradivarius violins with six newly made violins.\",\"D\":\"Stradivarius violins, crafted in Cremona between 1666 and 1737, were compared with newly made violins in a study involving 137 listeners.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"B\"},{\"id\":\"sat11-rw-m2-q32\",\"number\":32,\"type\":\"rhetorical_synthesis\",\"passage\":\"While researching bioluminescent organisms, a student has taken the following notes:\\n- Dinoflagellates are single-celled marine plankton.\\n- Some species, such as Noctiluca scintillans, emit blue-green light when disturbed.\\n- The light is produced through a chemical reaction involving luciferin and the enzyme luciferase.\\n- Researchers hypothesize that this flash startles predators, giving the dinoflagellate time to escape.\\n- The phenomenon is often called the 'burglar alarm' defense.\",\"prompt\":\"The student wants to explain the proposed function of bioluminescence in Noctiluca scintillans. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"Noctiluca scintillans is a single-celled marine plankton that produces a blue-green light through a chemical reaction.\",\"B\":\"The light produced by Noctiluca scintillans is generated by a chemical reaction involving luciferin and luciferase.\",\"C\":\"When disturbed, Noctiluca scintillans emits a flash of light that researchers believe startles predators, allowing the dinoflagellate to escape—a mechanism known as the 'burglar alarm' defense.\",\"D\":\"Bioluminescent dinoflagellates, including Noctiluca scintillans, are single-celled organisms that emit blue-green light when disturbed by passing predators.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"C\"},{\"id\":\"sat11-rw-m2-q33\",\"number\":33,\"type\":\"rhetorical_synthesis\",\"passage\":\"While researching the Harlem Renaissance, a student has taken the following notes:\\n- Zora Neale Hurston (1891-1960) was an American author and anthropologist.\\n- Her novel Their Eyes Were Watching God was published in 1937.\\n- The novel received mixed reviews upon its release.\\n- Richard Wright criticized it for lacking a clear political agenda.\\n- The novel was largely forgotten until Alice Walker rediscovered it in the 1970s.\\n- It is now considered a foundational work of African American literature.\",\"prompt\":\"The student wants to emphasize a contrast between the novel's initial reception and its current status. Which choice most effectively uses relevant information from the notes to accomplish this goal?\",\"choices\":{\"A\":\"Zora Neale Hurston's Their Eyes Were Watching God, published in 1937, received mixed reviews upon its release, though Richard Wright criticized it for lacking a clear political agenda.\",\"B\":\"Their Eyes Were Watching God, written by Zora Neale Hurston and published in 1937, was rediscovered by Alice Walker in the 1970s.\",\"C\":\"Although Alice Walker rediscovered Their Eyes Were Watching God in the 1970s, the novel had initially been criticized by Richard Wright.\",\"D\":\"Though initially met with mixed reviews and largely forgotten after its 1937 publication, Zora Neale Hurston's Their Eyes Were Watching God is now regarded as a foundational work of African American literature.\"},\"answer_format\":\"mcq\",\"correct_answer\":\"D\"}]" }, "status": "ok" } ], "succeeded": 14 }, "gen-math": { "count": 10, "costUsd": 0, "failed": 0, "items": [ { "costUsd": 0, "model": "gemini-2.5-pro", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 7 SAT Math questions (of which 2 are student-produced-response and the rest are MCQ).\nType: algebra | Test number: 11 | Module: 1 | First question number: 1\n\nMANDATORY: Your JSON array MUST contain exactly 7 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-math-m1-qN\" where N is the sequential question number starting at 1.\nFor example, the first question in this batch has id \"sat11-math-m1-q1\".\n\nMCQ Schema: {\"id\":\"sat11-math-m1-qN\",\"number\":N,\"type\":\"algebra\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\nSPR Schema: {\"id\":\"sat11-math-m1-qN\",\"number\":N,\"type\":\"algebra\",\"prompt\":\"...\",\"answer_format\":\"student_produced\",\"correct_answer\":\"numeric_value\"}\nPlace SPR questions at the END of the array.\n\nFollow instructions for your type:\n\n[algebra] (MCQ + 2 SPR)\n- Topics: linear equations (1 var, 2 vars), systems of 2 linear equations, linear inequalities, absolute value\n- Mix word problems and abstract algebra. Use $...$ notation.\n- MCQ distractors: sign errors, arithmetic mistakes, setup errors, partial answers\n- SPR answers: clean integers or simple fractions\n\n[functions_analysis] (MCQ + 2 SPR)\n- Topics: function evaluation, domain/range, transformations (f(x)+k, f(x-h)), graphical analysis, composition f(g(x)), piecewise\n- At least one conceptual insight question (e.g., (p,q) and (q,p) on same graph)\n- Describe graphs in text. All computations yield clean results.\n\n[geometry_trigonometry] (MCQ + 2 SPR)\n- Topics: circle equations, Pythagorean theorem, special right triangles, coordinate geometry, trig ratios, volume/surface area, angle relationships\n- Optional \"visual\" field: {\"type\":\"coordinate_plane\",\"description\":\"...\"}\n- Radical answers OK (e.g., $5\\sqrt{13}$). Multi-step reasoning required.\n\n[statistics_probability] (MCQ + 2 SPR)\n- Topics: mean/median/mode, standard deviation (conceptual), probability (conditional from tables), data interpretation, scatter plots/line of best fit, sampling/bias\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Distractors: P(A|B) vs P(B|A) confusion, mean vs median, causal from correlational\n\n[advanced_math] (MCQ + 2 SPR)\n- Topics: quadratics (factoring, discriminant), polynomials (zeros, end behavior), exponentials (growth/decay), rational expressions, radical expressions, non-linear systems\n- Hardest questions on the test. Multi-step reasoning required. Elegant algebraic insight preferred.\n- All expressions must be factorable with integer coefficients. SPR answers are clean numbers.\n\nVERIFY EVERY ANSWER: Solve each problem step-by-step before finalizing correct_answer.\nOutput ONLY a valid JSON array of EXACTLY 7 objects. No markdown. Count them before outputting.", "text": "```json\n[\n {\n \"id\": \"sat11-math-m1-q1\",\n \"number\": 1,\n \"type\": \"algebra\",\n \"prompt\": \"A taxi service charges a $2.50 pickup fee plus $1.75 per mile. If a passenger's total fare was $20.00, how many miles did the taxi travel?\",\n \"choices\": {\n \"A\": \"8\",\n \"B\": \"10\",\n \"C\": \"11.43\",\n \"D\": \"12.86\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"B\"\n },\n {\n \"id\": \"sat11-math-m1-q2\",\n \"number\": 2,\n \"type\": \"algebra\",\n \"prompt\": \"A small theater sells adult tickets for $12 each and child tickets for $8 each. If a total of 150 tickets were sold for a total of $1,560, how many adult tickets were sold?\",\n \"choices\": {\n \"A\": \"60\",\n \"B\": \"75\",\n \"C\": \"90\",\n \"D\": \"100\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m1-q3\",\n \"number\": 3,\n \"type\": \"algebra\",\n \"prompt\": \"Which of the following inequalities is equivalent to $7 - 2x > 15$?\",\n \"choices\": {\n \"A\": \"$x < -4$\",\n \"B\": \"$x > -4$\",\n \"C\": \"$x < 4$\",\n \"D\": \"$x > 4$\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\"\n },\n {\n \"id\": \"sat11-math-m1-q4\",\n \"number\": 4,\n \"type\": \"algebra\",\n \"prompt\": \"If $|\\frac{x}{2} - 3| = 4$, what is a possible value of $x$?\",\n \"choices\": {\n \"A\": \"2\",\n \"B\": \"7\",\n \"C\": \"-1\",\n \"D\": \"14\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"D\"\n },\n {\n \"id\": \"sat11-math-m1-q5\",\n \"number\": 5,\n \"type\": \"algebra\",\n \"prompt\": \"If $3x + 2y = 18$ and $2x + 3y = 17$, what is the value of $x+y$?\",\n \"choices\": {\n \"A\": \"1\",\n \"B\": \"4\",\n \"C\": \"7\",\n \"D\": \"35\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m1-q6\",\n \"number\": 6,\n \"type\": \"algebra\",\n \"prompt\": \"A rental company charges $40 for the first day and $25 for each additional day to rent a machine. If a rental bill for a machine was $215, for how many total days was it rented?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": \"8\"\n },\n {\n \"id\": \"sat11-math-m1-q7\",\n \"number\": 7,\n \"type\": \"algebra\",\n \"prompt\": \"The inequality $|n - 7| \\le 2$ describes the range of possible values for a variable $n$. If $n$ must be an integer, what is the largest possible value of $n$?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": \"9\"\n }\n]\n```" }, { "costUsd": 0, "model": "gemini-2.5-pro", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Math questions (of which 1 are student-produced-response and the rest are MCQ).\nType: functions_analysis | Test number: 11 | Module: 1 | First question number: 8\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-math-m1-qN\" where N is the sequential question number starting at 8.\nFor example, the first question in this batch has id \"sat11-math-m1-q8\".\n\nMCQ Schema: {\"id\":\"sat11-math-m1-qN\",\"number\":N,\"type\":\"functions_analysis\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\nSPR Schema: {\"id\":\"sat11-math-m1-qN\",\"number\":N,\"type\":\"functions_analysis\",\"prompt\":\"...\",\"answer_format\":\"student_produced\",\"correct_answer\":\"numeric_value\"}\nPlace SPR questions at the END of the array.\n\nFollow instructions for your type:\n\n[algebra] (MCQ + 1 SPR)\n- Topics: linear equations (1 var, 2 vars), systems of 2 linear equations, linear inequalities, absolute value\n- Mix word problems and abstract algebra. Use $...$ notation.\n- MCQ distractors: sign errors, arithmetic mistakes, setup errors, partial answers\n- SPR answers: clean integers or simple fractions\n\n[functions_analysis] (MCQ + 1 SPR)\n- Topics: function evaluation, domain/range, transformations (f(x)+k, f(x-h)), graphical analysis, composition f(g(x)), piecewise\n- At least one conceptual insight question (e.g., (p,q) and (q,p) on same graph)\n- Describe graphs in text. All computations yield clean results.\n\n[geometry_trigonometry] (MCQ + 1 SPR)\n- Topics: circle equations, Pythagorean theorem, special right triangles, coordinate geometry, trig ratios, volume/surface area, angle relationships\n- Optional \"visual\" field: {\"type\":\"coordinate_plane\",\"description\":\"...\"}\n- Radical answers OK (e.g., $5\\sqrt{13}$). Multi-step reasoning required.\n\n[statistics_probability] (MCQ + 1 SPR)\n- Topics: mean/median/mode, standard deviation (conceptual), probability (conditional from tables), data interpretation, scatter plots/line of best fit, sampling/bias\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Distractors: P(A|B) vs P(B|A) confusion, mean vs median, causal from correlational\n\n[advanced_math] (MCQ + 1 SPR)\n- Topics: quadratics (factoring, discriminant), polynomials (zeros, end behavior), exponentials (growth/decay), rational expressions, radical expressions, non-linear systems\n- Hardest questions on the test. Multi-step reasoning required. Elegant algebraic insight preferred.\n- All expressions must be factorable with integer coefficients. SPR answers are clean numbers.\n\nVERIFY EVERY ANSWER: Solve each problem step-by-step before finalizing correct_answer.\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "```json\n[\n {\n \"id\": \"sat11-math-m1-q8\",\n \"number\": 8,\n \"type\": \"functions_analysis\",\n \"prompt\": \"For the function $f$ defined by $f(x) = 3x^2 - 5x + 1$, what is the value of $f(-2)$?\",\n \"choices\": {\n \"A\": \"3\",\n \"B\": \"-1\",\n \"C\": \"23\",\n \"D\": \"-21\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m1-q9\",\n \"number\": 9,\n \"type\": \"functions_analysis\",\n \"prompt\": \"The graph of the function $g$ in the $xy$-plane contains the point $(2, 5)$. Which of the following points must be on the graph of the function $h$ defined by $h(x) = g(x-3) + 4$?\",\n \"choices\": {\n \"A\": \"(-1, 9)\",\n \"B\": \"(5, 9)\",\n \"C\": \"(5, 1)\",\n \"D\": \"(-1, 1)\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"B\"\n },\n {\n \"id\": \"sat11-math-m1-q10\",\n \"number\": 10,\n \"type\": \"functions_analysis\",\n \"prompt\": \"Let the functions $f$ and $g$ be defined by $f(x) = x^2 + 4$ and $g(x) = \\\\sqrt{x-1}$. What is the domain of the composite function $h(x) = f(g(x))$?\",\n \"choices\": {\n \"A\": \"x \\\\ge 1\",\n \"B\": \"x \\\\ge -4\",\n \"C\": \"x \\\\le 1\",\n \"D\": \"All real numbers\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\"\n },\n {\n \"id\": \"sat11-math-m1-q11\",\n \"number\": 11,\n \"type\": \"functions_analysis\",\n \"prompt\": \"A function $f$ is defined as follows:\\n$f(x) = \\\\begin{cases} 2x + 7, & \\\\text{if } x < 3 \\\\\\\\ x^2 - 1, & \\\\text{if } x \\\\ge 3 \\\\end{cases}$\\nWhat is the value of $f(3)$?\",\n \"choices\": {\n \"A\": \"13\",\n \"B\": \"10\",\n \"C\": \"9\",\n \"D\": \"8\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"D\"\n },\n {\n \"id\": \"sat11-math-m1-q12\",\n \"number\": 12,\n \"type\": \"functions_analysis\",\n \"prompt\": \"The functions $g$ and $h$ are defined by $g(x) = 2x - 5$ and $h(x) = x^2 + 3$. What is the value of $g(h(2))$?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": \"9\"\n }\n]\n```" }, { "costUsd": 0, "model": "gemini-2.5-pro", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Math questions (of which 1 are student-produced-response and the rest are MCQ).\nType: geometry_trigonometry | Test number: 11 | Module: 1 | First question number: 13\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-math-m1-qN\" where N is the sequential question number starting at 13.\nFor example, the first question in this batch has id \"sat11-math-m1-q13\".\n\nMCQ Schema: {\"id\":\"sat11-math-m1-qN\",\"number\":N,\"type\":\"geometry_trigonometry\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\nSPR Schema: {\"id\":\"sat11-math-m1-qN\",\"number\":N,\"type\":\"geometry_trigonometry\",\"prompt\":\"...\",\"answer_format\":\"student_produced\",\"correct_answer\":\"numeric_value\"}\nPlace SPR questions at the END of the array.\n\nFollow instructions for your type:\n\n[algebra] (MCQ + 1 SPR)\n- Topics: linear equations (1 var, 2 vars), systems of 2 linear equations, linear inequalities, absolute value\n- Mix word problems and abstract algebra. Use $...$ notation.\n- MCQ distractors: sign errors, arithmetic mistakes, setup errors, partial answers\n- SPR answers: clean integers or simple fractions\n\n[functions_analysis] (MCQ + 1 SPR)\n- Topics: function evaluation, domain/range, transformations (f(x)+k, f(x-h)), graphical analysis, composition f(g(x)), piecewise\n- At least one conceptual insight question (e.g., (p,q) and (q,p) on same graph)\n- Describe graphs in text. All computations yield clean results.\n\n[geometry_trigonometry] (MCQ + 1 SPR)\n- Topics: circle equations, Pythagorean theorem, special right triangles, coordinate geometry, trig ratios, volume/surface area, angle relationships\n- Optional \"visual\" field: {\"type\":\"coordinate_plane\",\"description\":\"...\"}\n- Radical answers OK (e.g., $5\\sqrt{13}$). Multi-step reasoning required.\n\n[statistics_probability] (MCQ + 1 SPR)\n- Topics: mean/median/mode, standard deviation (conceptual), probability (conditional from tables), data interpretation, scatter plots/line of best fit, sampling/bias\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Distractors: P(A|B) vs P(B|A) confusion, mean vs median, causal from correlational\n\n[advanced_math] (MCQ + 1 SPR)\n- Topics: quadratics (factoring, discriminant), polynomials (zeros, end behavior), exponentials (growth/decay), rational expressions, radical expressions, non-linear systems\n- Hardest questions on the test. Multi-step reasoning required. Elegant algebraic insight preferred.\n- All expressions must be factorable with integer coefficients. SPR answers are clean numbers.\n\nVERIFY EVERY ANSWER: Solve each problem step-by-step before finalizing correct_answer.\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "```json\n[\n {\n \"id\": \"sat11-math-m1-q13\",\n \"number\": 13,\n \"type\": \"geometry_trigonometry\",\n \"prompt\": \"In the xy-plane, a circle has a center at $(3, -5)$ and a radius of 7. Which of the following is the equation of the circle?\",\n \"choices\": {\n \"A\": \"$(x+3)^2 + (y-5)^2 = 7$\",\n \"B\": \"$(x-3)^2 + (y+5)^2 = 49$\",\n \"C\": \"$(x-3)^2 + (y+5)^2 = 7$\",\n \"D\": \"$(x+3)^2 + (y-5)^2 = 49$\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"B\"\n },\n {\n \"id\": \"sat11-math-m1-q14\",\n \"number\": 14,\n \"type\": \"geometry_trigonometry\",\n \"prompt\": \"In right triangle ABC, the right angle is at vertex C. If the length of side AC is 8 and the length of side BC is 15, what is the value of $\\\\cos(A)$?\",\n \"choices\": {\n \"A\": \"$8/17$\",\n \"B\": \"$15/17$\",\n \"C\": \"$8/15$\",\n \"D\": \"$17/8$\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\"\n },\n {\n \"id\": \"sat11-math-m1-q15\",\n \"number\": 15,\n \"type\": \"geometry_trigonometry\",\n \"prompt\": \"Two parallel lines are intersected by a transversal line. The measure of one of the resulting interior angles is $(3x + 10)^\\\\circ$ and the measure of its corresponding consecutive interior angle is $(5x + 30)^\\\\circ$. What is the value of $x$?\",\n \"choices\": {\n \"A\": \"10\",\n \"B\": \"17.5\",\n \"C\": \"20\",\n \"D\": \"40\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m1-q16\",\n \"number\": 16,\n \"type\": \"geometry_trigonometry\",\n \"prompt\": \"In right triangle $XYZ$, the measure of angle $Y$ is $90^\\\\circ$ and the measure of angle $X$ is $60^\\\\circ$. If side $XY$ (the side adjacent to angle $X$) has a length of 5, what is the length of the hypotenuse, $XZ$?\",\n \"choices\": {\n \"A\": \"$5\\\\sqrt{2}$\",\n \"B\": \"$5\\\\sqrt{3}$\",\n \"C\": \"$10\\\\sqrt{3}/3$\",\n \"D\": \"10\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"D\"\n },\n {\n \"id\": \"sat11-math-m1-q17\",\n \"number\": 17,\n \"type\": \"geometry_trigonometry\",\n \"prompt\": \"A rectangular prism has a height of 4 cm, a width of 3 cm, and a length of 12 cm. What is the length, in cm, of the space diagonal that connects two opposite vertices of the prism?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": \"13\"\n }\n]\n```" }, { "costUsd": 0, "model": "gemini-2.5-pro", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Math questions (of which 1 are student-produced-response and the rest are MCQ).\nType: statistics_probability | Test number: 11 | Module: 1 | First question number: 18\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-math-m1-qN\" where N is the sequential question number starting at 18.\nFor example, the first question in this batch has id \"sat11-math-m1-q18\".\n\nMCQ Schema: {\"id\":\"sat11-math-m1-qN\",\"number\":N,\"type\":\"statistics_probability\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\nSPR Schema: {\"id\":\"sat11-math-m1-qN\",\"number\":N,\"type\":\"statistics_probability\",\"prompt\":\"...\",\"answer_format\":\"student_produced\",\"correct_answer\":\"numeric_value\"}\nPlace SPR questions at the END of the array.\n\nFollow instructions for your type:\n\n[algebra] (MCQ + 1 SPR)\n- Topics: linear equations (1 var, 2 vars), systems of 2 linear equations, linear inequalities, absolute value\n- Mix word problems and abstract algebra. Use $...$ notation.\n- MCQ distractors: sign errors, arithmetic mistakes, setup errors, partial answers\n- SPR answers: clean integers or simple fractions\n\n[functions_analysis] (MCQ + 1 SPR)\n- Topics: function evaluation, domain/range, transformations (f(x)+k, f(x-h)), graphical analysis, composition f(g(x)), piecewise\n- At least one conceptual insight question (e.g., (p,q) and (q,p) on same graph)\n- Describe graphs in text. All computations yield clean results.\n\n[geometry_trigonometry] (MCQ + 1 SPR)\n- Topics: circle equations, Pythagorean theorem, special right triangles, coordinate geometry, trig ratios, volume/surface area, angle relationships\n- Optional \"visual\" field: {\"type\":\"coordinate_plane\",\"description\":\"...\"}\n- Radical answers OK (e.g., $5\\sqrt{13}$). Multi-step reasoning required.\n\n[statistics_probability] (MCQ + 1 SPR)\n- Topics: mean/median/mode, standard deviation (conceptual), probability (conditional from tables), data interpretation, scatter plots/line of best fit, sampling/bias\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Distractors: P(A|B) vs P(B|A) confusion, mean vs median, causal from correlational\n\n[advanced_math] (MCQ + 1 SPR)\n- Topics: quadratics (factoring, discriminant), polynomials (zeros, end behavior), exponentials (growth/decay), rational expressions, radical expressions, non-linear systems\n- Hardest questions on the test. Multi-step reasoning required. Elegant algebraic insight preferred.\n- All expressions must be factorable with integer coefficients. SPR answers are clean numbers.\n\nVERIFY EVERY ANSWER: Solve each problem step-by-step before finalizing correct_answer.\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "```json\n[\n {\n \"id\": \"sat11-math-m1-q18\",\n \"number\": 18,\n \"type\": \"statistics_probability\",\n \"prompt\": \"A community garden manager models the daily water usage for a plot of tomato plants. The line of best fit for the model is $y = 0.25x + 1.5$, where $y$ is the amount of water used in liters and $x$ is the average daily temperature in degrees Celsius. What is the best interpretation of the slope of this line?\",\n \"choices\": {\n \"A\": \"The initial amount of water used, regardless of temperature, is 0.25 liters.\",\n \"B\": \"For every 1-degree Celsius increase in temperature, the plants require an additional 1.5 liters of water.\",\n \"C\": \"For every 1-degree Celsius increase in average daily temperature, the predicted water usage increases by 0.25 liters.\",\n \"D\": \"For every 0.25-degree Celsius increase in temperature, the water usage increases by 1 liter.\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m1-q19\",\n \"number\": 19,\n \"type\": \"statistics_probability\",\n \"visual\": {\n \"type\": \"table\",\n \"title\": \"Instrument Played by Grade Level\",\n \"headers\": [\"\", \"Strings\", \"Woodwind\", \"Total\"],\n \"rows\": [\n [\"Grade 9\", 18, 22, 40],\n [\"Grade 10\", 24, 16, 40],\n [\"Total\", 42, 38, 80]\n ]\n },\n \"prompt\": \"The table above shows the distribution of 80 students in a music program by grade level and the type of instrument they play. If a student who plays a strings instrument is selected at random, what is the probability that the student is in Grade 9?\",\n \"choices\": {\n \"A\": \"$\\\\frac{18}{80}$\",\n \"B\": \"$\\\\frac{18}{40}$\",\n \"C\": \"$\\\\frac{40}{80}$\",\n \"D\": \"$\\\\frac{18}{42}$\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"D\"\n },\n {\n \"id\": \"sat11-math-m1-q20\",\n \"number\": 20,\n \"type\": \"statistics_probability\",\n \"prompt\": \"The list shows the number of points scored by a basketball player in each of the last 7 games: 15, 21, 18, 24, 19, 35, 17. Which statement accurately compares the mean and median of these scores?\",\n \"choices\": {\n \"A\": \"The mean is greater than the median.\",\n \"B\": \"The median is greater than the mean.\",\n \"C\": \"The mean and the median are equal.\",\n \"D\": \"The relationship cannot be determined from the information given.\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\"\n },\n {\n \"id\": \"sat11-math-m1-q21\",\n \"number\": 21,\n \"type\": \"statistics_probability\",\n \"visual\": {\n \"type\": \"table\",\n \"title\": \"City Employee Departments\",\n \"headers\": [\"Department\", \"Number of Employees\"],\n \"rows\": [\n [\"Public Works\", 450],\n [\"Parks & Recreation\", 200],\n [\"Administration\", 150],\n [\"Safety\", 700]\n ]\n },\n \"prompt\": \"A researcher wants to survey employees about workplace satisfaction in the city described in the table. Which sampling method is most likely to introduce bias?\",\n \"choices\": {\n \"A\": \"Using a random number generator to select 100 employees from a full alphabetical list of all city employees.\",\n \"B\": \"Leaving survey forms at the entrance of the Safety department building and using the first 100 that are returned.\",\n \"C\": \"Randomly selecting 50 employees from the Public Works department.\",\n \"D\": \"Creating a stratified sample by selecting employees from each department in proportion to the department's size.\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"B\"\n },\n {\n \"id\": \"sat11-math-m1-q22\",\n \"number\": 22,\n \"type\": \"statistics_probability\",\n \"prompt\": \"A dataset of 14 numbers has a mean of 68. A fifteenth number is added to the dataset, and the new mean is 70. What is the value of the number that was added to the dataset?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": 98\n }\n]\n```" }, { "costUsd": 0, "model": "gemini-2.5-pro", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Math questions (of which 2 are student-produced-response and the rest are MCQ).\nType: advanced_math | Test number: 11 | Module: 1 | First question number: 23\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-math-m1-qN\" where N is the sequential question number starting at 23.\nFor example, the first question in this batch has id \"sat11-math-m1-q23\".\n\nMCQ Schema: {\"id\":\"sat11-math-m1-qN\",\"number\":N,\"type\":\"advanced_math\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\nSPR Schema: {\"id\":\"sat11-math-m1-qN\",\"number\":N,\"type\":\"advanced_math\",\"prompt\":\"...\",\"answer_format\":\"student_produced\",\"correct_answer\":\"numeric_value\"}\nPlace SPR questions at the END of the array.\n\nFollow instructions for your type:\n\n[algebra] (MCQ + 2 SPR)\n- Topics: linear equations (1 var, 2 vars), systems of 2 linear equations, linear inequalities, absolute value\n- Mix word problems and abstract algebra. Use $...$ notation.\n- MCQ distractors: sign errors, arithmetic mistakes, setup errors, partial answers\n- SPR answers: clean integers or simple fractions\n\n[functions_analysis] (MCQ + 2 SPR)\n- Topics: function evaluation, domain/range, transformations (f(x)+k, f(x-h)), graphical analysis, composition f(g(x)), piecewise\n- At least one conceptual insight question (e.g., (p,q) and (q,p) on same graph)\n- Describe graphs in text. All computations yield clean results.\n\n[geometry_trigonometry] (MCQ + 2 SPR)\n- Topics: circle equations, Pythagorean theorem, special right triangles, coordinate geometry, trig ratios, volume/surface area, angle relationships\n- Optional \"visual\" field: {\"type\":\"coordinate_plane\",\"description\":\"...\"}\n- Radical answers OK (e.g., $5\\sqrt{13}$). Multi-step reasoning required.\n\n[statistics_probability] (MCQ + 2 SPR)\n- Topics: mean/median/mode, standard deviation (conceptual), probability (conditional from tables), data interpretation, scatter plots/line of best fit, sampling/bias\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Distractors: P(A|B) vs P(B|A) confusion, mean vs median, causal from correlational\n\n[advanced_math] (MCQ + 2 SPR)\n- Topics: quadratics (factoring, discriminant), polynomials (zeros, end behavior), exponentials (growth/decay), rational expressions, radical expressions, non-linear systems\n- Hardest questions on the test. Multi-step reasoning required. Elegant algebraic insight preferred.\n- All expressions must be factorable with integer coefficients. SPR answers are clean numbers.\n\nVERIFY EVERY ANSWER: Solve each problem step-by-step before finalizing correct_answer.\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "```json\n[\n {\n \"id\": \"sat11-math-m1-q23\",\n \"number\": 23,\n \"type\": \"advanced_math\",\n \"prompt\": \"In the xy-plane, a parabola with the equation $y = x^2 + bx + c$ and a line with the equation $y = -x + k$ intersect at exactly two points, $(2, 5)$ and $(5, 2)$. What is the value of $b$?\",\n \"choices\": {\n \"A\": \"-8\",\n \"B\": \"-6\",\n \"C\": \"2\",\n \"D\": \"7\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"B\"\n },\n {\n \"id\": \"sat11-math-m1-q24\",\n \"number\": 24,\n \"type\": \"advanced_math\",\n \"prompt\": \"Which of the following expressions is equivalent to $\\\\frac{2x^2-11x+10}{x-4}$ for $x>4$?\",\n \"choices\": {\n \"A\": \"$2x-3 - \\\\frac{2}{x-4}$\",\n \"B\": \"$2x+3 + \\\\frac{22}{x-4}$\",\n \"C\": \"$2x-19 + \\\\frac{86}{x-4}$\",\n \"D\": \"$2x-3 + \\\\frac{22}{x-4}$\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\"\n },\n {\n \"id\": \"sat11-math-m1-q25\",\n \"number\": 25,\n \"type\": \"advanced_math\",\n \"prompt\": \"The function $f$ is defined by $f(x) = k(x+a)(x+b)$, where $k$, $a$, and $b$ are constants. The graph of $y=f(x)$ in the xy-plane is a parabola with x-intercepts at $-3$ and $5$. The vertex of the parabola is $(1, -32)$. What is the value of $k$?\",\n \"choices\": {\n \"A\": \"-2\",\n \"B\": \"1\",\n \"C\": \"2\",\n \"D\": \"8\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m1-q26\",\n \"number\": 26,\n \"type\": \"advanced_math\",\n \"prompt\": \"If $x > 0$ and $\\\\sqrt{3\\\\sqrt{x}} = \\\\sqrt[n]{3x}$ for all values of $x$, what is the value of $n$?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": 4\n },\n {\n \"id\": \"sat11-math-m1-q27\",\n \"number\": 27,\n \"type\": \"advanced_math\",\n \"prompt\": \"For the quadratic equation $kx^2 - 8x + 2 = 0$, the sum of the solutions is twice the product of the solutions. What is the value of $k$?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": 2\n }\n]\n```" }, { "costUsd": 0, "model": "gemini-2.5-pro", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 7 SAT Math questions (of which 2 are student-produced-response and the rest are MCQ).\nType: algebra | Test number: 11 | Module: 2 | First question number: 1\n\nMANDATORY: Your JSON array MUST contain exactly 7 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-math-m2-qN\" where N is the sequential question number starting at 1.\nFor example, the first question in this batch has id \"sat11-math-m2-q1\".\n\nMCQ Schema: {\"id\":\"sat11-math-m2-qN\",\"number\":N,\"type\":\"algebra\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\nSPR Schema: {\"id\":\"sat11-math-m2-qN\",\"number\":N,\"type\":\"algebra\",\"prompt\":\"...\",\"answer_format\":\"student_produced\",\"correct_answer\":\"numeric_value\"}\nPlace SPR questions at the END of the array.\n\nFollow instructions for your type:\n\n[algebra] (MCQ + 2 SPR)\n- Topics: linear equations (1 var, 2 vars), systems of 2 linear equations, linear inequalities, absolute value\n- Mix word problems and abstract algebra. Use $...$ notation.\n- MCQ distractors: sign errors, arithmetic mistakes, setup errors, partial answers\n- SPR answers: clean integers or simple fractions\n\n[functions_analysis] (MCQ + 2 SPR)\n- Topics: function evaluation, domain/range, transformations (f(x)+k, f(x-h)), graphical analysis, composition f(g(x)), piecewise\n- At least one conceptual insight question (e.g., (p,q) and (q,p) on same graph)\n- Describe graphs in text. All computations yield clean results.\n\n[geometry_trigonometry] (MCQ + 2 SPR)\n- Topics: circle equations, Pythagorean theorem, special right triangles, coordinate geometry, trig ratios, volume/surface area, angle relationships\n- Optional \"visual\" field: {\"type\":\"coordinate_plane\",\"description\":\"...\"}\n- Radical answers OK (e.g., $5\\sqrt{13}$). Multi-step reasoning required.\n\n[statistics_probability] (MCQ + 2 SPR)\n- Topics: mean/median/mode, standard deviation (conceptual), probability (conditional from tables), data interpretation, scatter plots/line of best fit, sampling/bias\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Distractors: P(A|B) vs P(B|A) confusion, mean vs median, causal from correlational\n\n[advanced_math] (MCQ + 2 SPR)\n- Topics: quadratics (factoring, discriminant), polynomials (zeros, end behavior), exponentials (growth/decay), rational expressions, radical expressions, non-linear systems\n- Hardest questions on the test. Multi-step reasoning required. Elegant algebraic insight preferred.\n- All expressions must be factorable with integer coefficients. SPR answers are clean numbers.\n\nVERIFY EVERY ANSWER: Solve each problem step-by-step before finalizing correct_answer.\nOutput ONLY a valid JSON array of EXACTLY 7 objects. No markdown. Count them before outputting.", "text": "```json\n[\n {\n \"id\": \"sat11-math-m2-q1\",\n \"number\": 1,\n \"type\": \"algebra\",\n \"prompt\": \"If $\\\\frac{1}{3}(6x + 15) - 4 = 9$, what is the value of $x$?\",\n \"choices\": {\n \"A\": \"-1\",\n \"B\": \"4\",\n \"C\": \"5\",\n \"D\": \"8\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"B\"\n },\n {\n \"id\": \"sat11-math-m2-q2\",\n \"number\": 2,\n \"type\": \"algebra\",\n \"prompt\": \"A gym offers two membership plans. Plan A has a one-time initiation fee of $50 and a monthly fee of $20. Plan B has no initiation fee and a monthly fee of $30. After how many months will the total cost of both plans be equal?\",\n \"choices\": {\n \"A\": \"1.5\",\n \"B\": \"2.5\",\n \"C\": \"5\",\n \"D\": \"10\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m2-q3\",\n \"number\": 3,\n \"type\": \"algebra\",\n \"prompt\": \"Which of the following is a possible value of $y$ if $-4(y - 2) > 12$?\",\n \"choices\": {\n \"A\": \"-2\",\n \"B\": \"-1\",\n \"C\": \"0\",\n \"D\": \"2\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\"\n },\n {\n \"id\": \"sat11-math-m2-q4\",\n \"number\": 4,\n \"type\": \"algebra\",\n \"prompt\": \"If $|2k - 3| = 7$, what is the sum of the two possible values for $k$?\",\n \"choices\": {\n \"A\": \"-2\",\n \"B\": \"0\",\n \"C\": \"3\",\n \"D\": \"5\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m2-q5\",\n \"number\": 5,\n \"type\": \"algebra\",\n \"prompt\": \"If $5x - 2y = 11$ and $x + 3y = 9$, what is the value of $x$?\",\n \"choices\": {\n \"A\": \"-1\",\n \"B\": \"2\",\n \"C\": \"2.5\",\n \"D\": \"3\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"D\"\n },\n {\n \"id\": \"sat11-math-m2-q6\",\n \"number\": 6,\n \"type\": \"algebra\",\n \"prompt\": \"A rental car company charges a flat daily fee plus a fee per mile driven. A customer who drove 100 miles was charged $80. Another customer drove 150 miles on the same day and was charged $95. What is the company's flat daily fee in dollars?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": 50\n },\n {\n \"id\": \"sat11-math-m2-q7\",\n \"number\": 7,\n \"type\": \"algebra\",\n \"prompt\": \"The inequality $|x - 5| \\\\leq 2$ defines a range of values for $x$. If $x$ must be an integer, what is the sum of all possible values of $x$?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": 25\n }\n]\n```" }, { "costUsd": 0, "model": "gemini-2.5-pro", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Math questions (of which 1 are student-produced-response and the rest are MCQ).\nType: functions_analysis | Test number: 11 | Module: 2 | First question number: 8\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-math-m2-qN\" where N is the sequential question number starting at 8.\nFor example, the first question in this batch has id \"sat11-math-m2-q8\".\n\nMCQ Schema: {\"id\":\"sat11-math-m2-qN\",\"number\":N,\"type\":\"functions_analysis\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\nSPR Schema: {\"id\":\"sat11-math-m2-qN\",\"number\":N,\"type\":\"functions_analysis\",\"prompt\":\"...\",\"answer_format\":\"student_produced\",\"correct_answer\":\"numeric_value\"}\nPlace SPR questions at the END of the array.\n\nFollow instructions for your type:\n\n[algebra] (MCQ + 1 SPR)\n- Topics: linear equations (1 var, 2 vars), systems of 2 linear equations, linear inequalities, absolute value\n- Mix word problems and abstract algebra. Use $...$ notation.\n- MCQ distractors: sign errors, arithmetic mistakes, setup errors, partial answers\n- SPR answers: clean integers or simple fractions\n\n[functions_analysis] (MCQ + 1 SPR)\n- Topics: function evaluation, domain/range, transformations (f(x)+k, f(x-h)), graphical analysis, composition f(g(x)), piecewise\n- At least one conceptual insight question (e.g., (p,q) and (q,p) on same graph)\n- Describe graphs in text. All computations yield clean results.\n\n[geometry_trigonometry] (MCQ + 1 SPR)\n- Topics: circle equations, Pythagorean theorem, special right triangles, coordinate geometry, trig ratios, volume/surface area, angle relationships\n- Optional \"visual\" field: {\"type\":\"coordinate_plane\",\"description\":\"...\"}\n- Radical answers OK (e.g., $5\\sqrt{13}$). Multi-step reasoning required.\n\n[statistics_probability] (MCQ + 1 SPR)\n- Topics: mean/median/mode, standard deviation (conceptual), probability (conditional from tables), data interpretation, scatter plots/line of best fit, sampling/bias\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Distractors: P(A|B) vs P(B|A) confusion, mean vs median, causal from correlational\n\n[advanced_math] (MCQ + 1 SPR)\n- Topics: quadratics (factoring, discriminant), polynomials (zeros, end behavior), exponentials (growth/decay), rational expressions, radical expressions, non-linear systems\n- Hardest questions on the test. Multi-step reasoning required. Elegant algebraic insight preferred.\n- All expressions must be factorable with integer coefficients. SPR answers are clean numbers.\n\nVERIFY EVERY ANSWER: Solve each problem step-by-step before finalizing correct_answer.\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "```json\n[\n {\n \"id\": \"sat11-math-m2-q8\",\n \"number\": 8,\n \"type\": \"functions_analysis\",\n \"prompt\": \"Let the function $h$ be defined by $h(x) = x^2 - 3x$. If $h(a+2) = 4$ for some constant $a$, what is one possible value of $a$?\",\n \"choices\": {\n \"A\": \"2\",\n \"B\": \"-1\",\n \"C\": \"0\",\n \"D\": \"4\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\"\n },\n {\n \"id\": \"sat11-math-m2-q9\",\n \"number\": 9,\n \"type\": \"functions_analysis\",\n \"prompt\": \"The graph of a function $f$ in the $xy$-plane passes through the point $(0, 3)$. A second function, $g$, is defined by the transformation $g(x) = f(x-2) + 1$. Which of the following points must be on the graph of $g$?\",\n \"choices\": {\n \"A\": \"(-2, 4)\",\n \"B\": \"(2, 4)\",\n \"C\": \"(2, 2)\",\n \"D\": \"(-2, 2)\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"B\"\n },\n {\n \"id\": \"sat11-math-m2-q10\",\n \"number\": 10,\n \"type\": \"functions_analysis\",\n \"prompt\": \"A function $f$ is defined as $f(x) = \\\\begin{cases} 2x + 7 & \\\\text{if } x \\\\le 1 \\\\\\\\ x^2 - 4 & \\\\text{if } x > 1 \\\\end{cases}$. What is the value of $f(-3) + f(3)$?\",\n \"choices\": {\n \"A\": \"14\",\n \"B\": \"10\",\n \"C\": \"6\",\n \"D\": \"4\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m2-q11\",\n \"number\": 11,\n \"type\": \"functions_analysis\",\n \"prompt\": \"The function $g$ is defined by $g(x) = \\\\frac{x+3}{\\\\sqrt{4-x}}$. For the function's output to be a real number, which of the following inequalities describes all possible values of the input $x$?\",\n \"choices\": {\n \"A\": \"$x > 4$\",\n \"B\": \"$x \\\\ge 4$\",\n \"C\": \"$x \\\\le 4$\",\n \"D\": \"$x < 4$\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"D\"\n },\n {\n \"id\": \"sat11-math-m2-q12\",\n \"number\": 12,\n \"type\": \"functions_analysis\",\n \"prompt\": \"Let the functions $f$ and $g$ be defined by $f(x) = 2x - 5$ and $g(x) = x^2 + k$, where $k$ is a constant. If $f(g(1)) = -3$, what is the value of $k$?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": \"0\"\n }\n]\n```" }, { "costUsd": 0, "model": "gemini-2.5-pro", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Math questions (of which 1 are student-produced-response and the rest are MCQ).\nType: geometry_trigonometry | Test number: 11 | Module: 2 | First question number: 13\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-math-m2-qN\" where N is the sequential question number starting at 13.\nFor example, the first question in this batch has id \"sat11-math-m2-q13\".\n\nMCQ Schema: {\"id\":\"sat11-math-m2-qN\",\"number\":N,\"type\":\"geometry_trigonometry\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\nSPR Schema: {\"id\":\"sat11-math-m2-qN\",\"number\":N,\"type\":\"geometry_trigonometry\",\"prompt\":\"...\",\"answer_format\":\"student_produced\",\"correct_answer\":\"numeric_value\"}\nPlace SPR questions at the END of the array.\n\nFollow instructions for your type:\n\n[algebra] (MCQ + 1 SPR)\n- Topics: linear equations (1 var, 2 vars), systems of 2 linear equations, linear inequalities, absolute value\n- Mix word problems and abstract algebra. Use $...$ notation.\n- MCQ distractors: sign errors, arithmetic mistakes, setup errors, partial answers\n- SPR answers: clean integers or simple fractions\n\n[functions_analysis] (MCQ + 1 SPR)\n- Topics: function evaluation, domain/range, transformations (f(x)+k, f(x-h)), graphical analysis, composition f(g(x)), piecewise\n- At least one conceptual insight question (e.g., (p,q) and (q,p) on same graph)\n- Describe graphs in text. All computations yield clean results.\n\n[geometry_trigonometry] (MCQ + 1 SPR)\n- Topics: circle equations, Pythagorean theorem, special right triangles, coordinate geometry, trig ratios, volume/surface area, angle relationships\n- Optional \"visual\" field: {\"type\":\"coordinate_plane\",\"description\":\"...\"}\n- Radical answers OK (e.g., $5\\sqrt{13}$). Multi-step reasoning required.\n\n[statistics_probability] (MCQ + 1 SPR)\n- Topics: mean/median/mode, standard deviation (conceptual), probability (conditional from tables), data interpretation, scatter plots/line of best fit, sampling/bias\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Distractors: P(A|B) vs P(B|A) confusion, mean vs median, causal from correlational\n\n[advanced_math] (MCQ + 1 SPR)\n- Topics: quadratics (factoring, discriminant), polynomials (zeros, end behavior), exponentials (growth/decay), rational expressions, radical expressions, non-linear systems\n- Hardest questions on the test. Multi-step reasoning required. Elegant algebraic insight preferred.\n- All expressions must be factorable with integer coefficients. SPR answers are clean numbers.\n\nVERIFY EVERY ANSWER: Solve each problem step-by-step before finalizing correct_answer.\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "```json\n[\n {\n \"id\": \"sat11-math-m2-q13\",\n \"number\": 13,\n \"type\": \"geometry_trigonometry\",\n \"prompt\": \"In the xy-plane, a circle has a diameter with endpoints at $(2, -5)$ and $(10, 1)$. Which of the following is an equation of the circle?\",\n \"choices\": {\n \"A\": \"$(x - 6)^2 + (y + 2)^2 = 100$\",\n \"B\": \"$(x + 6)^2 + (y - 2)^2 = 25$\",\n \"C\": \"$(x - 4)^2 + (y + 2)^2 = 50$\",\n \"D\": \"$(x - 6)^2 + (y + 2)^2 = 25$\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"D\"\n },\n {\n \"id\": \"sat11-math-m2-q14\",\n \"number\": 14,\n \"type\": \"geometry_trigonometry\",\n \"prompt\": \"In right triangle $ABC$, the measure of angle $C$ is $90^\\\\circ$. If the length of side $AB$ is 13 and the length of side $BC$ is 5, what is the value of $\\\\tan(A)$?\",\n \"choices\": {\n \"A\": \"$\\\\frac{5}{13}$\",\n \"B\": \"$\\\\frac{12}{13}$\",\n \"C\": \"$\\\\frac{5}{12}$\",\n \"D\": \"$\\\\frac{12}{5}$\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m2-q15\",\n \"number\": 15,\n \"type\": \"geometry_trigonometry\",\n \"prompt\": \"A right circular cylinder and a right circular cone have the same radius and equal volumes. If the height of the cylinder is 12 inches, what is the height of the cone, in inches?\",\n \"choices\": {\n \"A\": \"36\",\n \"B\": \"12\",\n \"C\": \"4\",\n \"D\": \"24\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\"\n },\n {\n \"id\": \"sat11-math-m2-q16\",\n \"number\": 16,\n \"type\": \"geometry_trigonometry\",\n \"prompt\": \"In the xy-plane, $\\\\triangle PQR$ has vertices $P(0,0)$, $Q(6,0)$, and $R(3,4)$. If $\\\\triangle STU$ is similar to $\\\\triangle PQR$, with vertex $S$ at $(0,0)$ and vertex $T$ at $(9,0)$, what is the y-coordinate of vertex $U$?\",\n \"choices\": {\n \"A\": \"$\\\\frac{8}{3}$\",\n \"B\": \"6\",\n \"C\": \"4\",\n \"D\": \"7\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"B\"\n },\n {\n \"id\": \"sat11-math-m2-q17\",\n \"number\": 17,\n \"type\": \"geometry_trigonometry\",\n \"prompt\": \"A regular hexagon is inscribed in a circle. What is the measure in degrees of the acute angle formed between a side of the hexagon and a line tangent to the circle at a shared vertex?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": 30\n }\n]\n```" }, { "costUsd": 0, "model": "gemini-2.5-pro", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Math questions (of which 1 are student-produced-response and the rest are MCQ).\nType: statistics_probability | Test number: 11 | Module: 2 | First question number: 18\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-math-m2-qN\" where N is the sequential question number starting at 18.\nFor example, the first question in this batch has id \"sat11-math-m2-q18\".\n\nMCQ Schema: {\"id\":\"sat11-math-m2-qN\",\"number\":N,\"type\":\"statistics_probability\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\nSPR Schema: {\"id\":\"sat11-math-m2-qN\",\"number\":N,\"type\":\"statistics_probability\",\"prompt\":\"...\",\"answer_format\":\"student_produced\",\"correct_answer\":\"numeric_value\"}\nPlace SPR questions at the END of the array.\n\nFollow instructions for your type:\n\n[algebra] (MCQ + 1 SPR)\n- Topics: linear equations (1 var, 2 vars), systems of 2 linear equations, linear inequalities, absolute value\n- Mix word problems and abstract algebra. Use $...$ notation.\n- MCQ distractors: sign errors, arithmetic mistakes, setup errors, partial answers\n- SPR answers: clean integers or simple fractions\n\n[functions_analysis] (MCQ + 1 SPR)\n- Topics: function evaluation, domain/range, transformations (f(x)+k, f(x-h)), graphical analysis, composition f(g(x)), piecewise\n- At least one conceptual insight question (e.g., (p,q) and (q,p) on same graph)\n- Describe graphs in text. All computations yield clean results.\n\n[geometry_trigonometry] (MCQ + 1 SPR)\n- Topics: circle equations, Pythagorean theorem, special right triangles, coordinate geometry, trig ratios, volume/surface area, angle relationships\n- Optional \"visual\" field: {\"type\":\"coordinate_plane\",\"description\":\"...\"}\n- Radical answers OK (e.g., $5\\sqrt{13}$). Multi-step reasoning required.\n\n[statistics_probability] (MCQ + 1 SPR)\n- Topics: mean/median/mode, standard deviation (conceptual), probability (conditional from tables), data interpretation, scatter plots/line of best fit, sampling/bias\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Distractors: P(A|B) vs P(B|A) confusion, mean vs median, causal from correlational\n\n[advanced_math] (MCQ + 1 SPR)\n- Topics: quadratics (factoring, discriminant), polynomials (zeros, end behavior), exponentials (growth/decay), rational expressions, radical expressions, non-linear systems\n- Hardest questions on the test. Multi-step reasoning required. Elegant algebraic insight preferred.\n- All expressions must be factorable with integer coefficients. SPR answers are clean numbers.\n\nVERIFY EVERY ANSWER: Solve each problem step-by-step before finalizing correct_answer.\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "```json\n[\n {\n \"id\": \"sat11-math-m2-q18\",\n \"number\": 18,\n \"type\": \"statistics_probability\",\n \"prompt\": \"The scatterplot above shows the relationship between two variables, $x$ and $y$. A line of best fit for the data is also shown. For the data point $(10, 55)$, the actual $y$-value is 55. What is the $y$-value predicted by the line of best fit when $x=10$, and what is the value of the corresponding residual?\",\n \"visual\": {\n \"type\": \"coordinate_plane\",\n \"description\": \"A scatterplot with about 15 points that show a strong, positive, linear association. The line of best fit is drawn through the points. The line passes through approximately (0, 10) and (20, 90). The point (10, 55) is clearly plotted slightly above the line of best fit.\"\n },\n \"choices\": {\n \"A\": \"Predicted value: 50; Residual: -5\",\n \"B\": \"Predicted value: 55; Residual: 0\",\n \"C\": \"Predicted value: 50; Residual: 5\",\n \"D\": \"Predicted value: 5; Residual: 50\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m2-q19\",\n \"number\": 19,\n \"type\": \"statistics_probability\",\n \"prompt\": \"The table above summarizes the results of a survey of 200 smartphone users on their preferred mobile operating system and their primary use for the phone. Based on the table, what is the probability that a user who primarily uses their phone for communication prefers OS Y?\",\n \"visual\": {\n \"type\": \"table\",\n \"title\": \"Smartphone User Survey\",\n \"headers\": [\"Primary Use\", \"OS X\", \"OS Y\", \"Total\"],\n \"rows\": [\n [\"Communication\", 60, 40, 100],\n [\"Entertainment\", 70, 30, 100],\n [\"Total\", 130, 70, 200]\n ]\n },\n \"choices\": {\n \"A\": \"0.20\",\n \"B\": \"0.35\",\n \"C\": \"0.57\",\n \"D\": \"0.40\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"D\"\n },\n {\n \"id\": \"sat11-math-m2-q20\",\n \"number\": 20,\n \"type\": \"statistics_probability\",\n \"prompt\": \"A small business has 11 employees. The annual salaries, in thousands of dollars, are: 45, 45, 48, 50, 52, 55, 58, 60, 65, 70, and 300. Which of the following statements best compares the mean and median of these salaries?\",\n \"choices\": {\n \"A\": \"The mean is greater than the median.\",\n \"B\": \"The median is greater than the mean.\",\n \"C\": \"The mean and the median are equal.\",\n \"D\": \"The relationship between the mean and median cannot be determined.\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\"\n },\n {\n \"id\": \"sat11-math-m2-q21\",\n \"number\": 21,\n \"type\": \"statistics_probability\",\n \"prompt\": \"A market researcher for a national coffee chain wants to determine the average amount of money a customer spends per visit. Which of the following sampling methods is most likely to introduce a significant bias into the study?\",\n \"choices\": {\n \"A\": \"Selecting 100 customers at random from the company's national loyalty program database.\",\n \"B\": \"Surveying every customer who visits a single, specific store location between 8 a.m. and 9 a.m. on a Monday morning.\",\n \"C\": \"Dividing the chain's stores into regions and randomly selecting 5 stores from each region, then surveying all customers at those stores for one day.\",\n \"D\": \"Using a computer to randomly select 200 transaction times from the previous week and surveying the customer who made each of those transactions.\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"B\"\n },\n {\n \"id\": \"sat11-math-m2-q22\",\n \"number\": 22,\n \"type\": \"statistics_probability\",\n \"prompt\": \"At a conference of 300 attendees, 180 are from North America and the rest are from Europe. If 60% of the North American attendees are scientists and 40% of the European attendees are scientists, what is the probability that an attendee selected at random is a scientist? Express your answer as a decimal.\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": \"0.52\"\n }\n]\n```" }, { "costUsd": 0, "model": "gemini-2.5-pro", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Math questions (of which 2 are student-produced-response and the rest are MCQ).\nType: advanced_math | Test number: 11 | Module: 2 | First question number: 23\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-math-m2-qN\" where N is the sequential question number starting at 23.\nFor example, the first question in this batch has id \"sat11-math-m2-q23\".\n\nMCQ Schema: {\"id\":\"sat11-math-m2-qN\",\"number\":N,\"type\":\"advanced_math\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\nSPR Schema: {\"id\":\"sat11-math-m2-qN\",\"number\":N,\"type\":\"advanced_math\",\"prompt\":\"...\",\"answer_format\":\"student_produced\",\"correct_answer\":\"numeric_value\"}\nPlace SPR questions at the END of the array.\n\nFollow instructions for your type:\n\n[algebra] (MCQ + 2 SPR)\n- Topics: linear equations (1 var, 2 vars), systems of 2 linear equations, linear inequalities, absolute value\n- Mix word problems and abstract algebra. Use $...$ notation.\n- MCQ distractors: sign errors, arithmetic mistakes, setup errors, partial answers\n- SPR answers: clean integers or simple fractions\n\n[functions_analysis] (MCQ + 2 SPR)\n- Topics: function evaluation, domain/range, transformations (f(x)+k, f(x-h)), graphical analysis, composition f(g(x)), piecewise\n- At least one conceptual insight question (e.g., (p,q) and (q,p) on same graph)\n- Describe graphs in text. All computations yield clean results.\n\n[geometry_trigonometry] (MCQ + 2 SPR)\n- Topics: circle equations, Pythagorean theorem, special right triangles, coordinate geometry, trig ratios, volume/surface area, angle relationships\n- Optional \"visual\" field: {\"type\":\"coordinate_plane\",\"description\":\"...\"}\n- Radical answers OK (e.g., $5\\sqrt{13}$). Multi-step reasoning required.\n\n[statistics_probability] (MCQ + 2 SPR)\n- Topics: mean/median/mode, standard deviation (conceptual), probability (conditional from tables), data interpretation, scatter plots/line of best fit, sampling/bias\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Distractors: P(A|B) vs P(B|A) confusion, mean vs median, causal from correlational\n\n[advanced_math] (MCQ + 2 SPR)\n- Topics: quadratics (factoring, discriminant), polynomials (zeros, end behavior), exponentials (growth/decay), rational expressions, radical expressions, non-linear systems\n- Hardest questions on the test. Multi-step reasoning required. Elegant algebraic insight preferred.\n- All expressions must be factorable with integer coefficients. SPR answers are clean numbers.\n\nVERIFY EVERY ANSWER: Solve each problem step-by-step before finalizing correct_answer.\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "```json\n[\n {\n \"id\": \"sat11-math-m2-q23\",\n \"number\": 23,\n \"type\": \"advanced_math\",\n \"prompt\": \"The parabola with equation $y = x^2 - 5x + 11$ intersects the line with equation $y = 3x - 4$ at two points, $(x_1, y_1)$ and $(x_2, y_2)$. What is the value of $x_1 + x_2$?\",\n \"choices\": {\n \"A\": \"8\",\n \"B\": \"15\",\n \"C\": \"2\",\n \"D\": \"-2\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\"\n },\n {\n \"id\": \"sat11-math-m2-q24\",\n \"number\": 24,\n \"type\": \"advanced_math\",\n \"prompt\": \"The expression $\\\\frac{3x^2+5x-2}{x^2-4}$ can be written as $A + \\\\frac{B}{x-2}$, where $A$ and $B$ are constants and $x \\\\neq \\\\pm 2$. What is the value of $B$?\",\n \"choices\": {\n \"A\": \"3\",\n \"B\": \"5\",\n \"C\": \"8.5\",\n \"D\": \"4.5\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"D\"\n },\n {\n \"id\": \"sat11-math-m2-q25\",\n \"number\": 25,\n \"type\": \"advanced_math\",\n \"prompt\": \"For a function defined by $f(x) = c \\cdot p^x$, where $c$ and $p$ are positive constants, $f(1)=15$ and $f(3)=135$. What is the value of $p$?\",\n \"choices\": {\n \"A\": \"9\",\n \"B\": \"5\",\n \"C\": \"3\",\n \"D\": \"60\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m2-q26\",\n \"number\": 26,\n \"type\": \"advanced_math\",\n \"prompt\": \"If $x > 0$ and $2\\sqrt{x+2} = x+1$, what is the value of $x$?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": 2\n },\n {\n \"id\": \"sat11-math-m2-q27\",\n \"number\": 27,\n \"type\": \"advanced_math\",\n \"prompt\": \"In the $xy$-plane, the graphs of the functions $f(x) = -x^2+3$ and $g(x) = x^2+bx-1$ intersect at exactly one point. What is the value of $b$?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": 4\n }\n]\n```" } ], "results": [ { "attempts": 1, "index": 0, "output": { "costUsd": 0, "model": "gemini-2.5-pro", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 7 SAT Math questions (of which 2 are student-produced-response and the rest are MCQ).\nType: algebra | Test number: 11 | Module: 1 | First question number: 1\n\nMANDATORY: Your JSON array MUST contain exactly 7 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-math-m1-qN\" where N is the sequential question number starting at 1.\nFor example, the first question in this batch has id \"sat11-math-m1-q1\".\n\nMCQ Schema: {\"id\":\"sat11-math-m1-qN\",\"number\":N,\"type\":\"algebra\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\nSPR Schema: {\"id\":\"sat11-math-m1-qN\",\"number\":N,\"type\":\"algebra\",\"prompt\":\"...\",\"answer_format\":\"student_produced\",\"correct_answer\":\"numeric_value\"}\nPlace SPR questions at the END of the array.\n\nFollow instructions for your type:\n\n[algebra] (MCQ + 2 SPR)\n- Topics: linear equations (1 var, 2 vars), systems of 2 linear equations, linear inequalities, absolute value\n- Mix word problems and abstract algebra. Use $...$ notation.\n- MCQ distractors: sign errors, arithmetic mistakes, setup errors, partial answers\n- SPR answers: clean integers or simple fractions\n\n[functions_analysis] (MCQ + 2 SPR)\n- Topics: function evaluation, domain/range, transformations (f(x)+k, f(x-h)), graphical analysis, composition f(g(x)), piecewise\n- At least one conceptual insight question (e.g., (p,q) and (q,p) on same graph)\n- Describe graphs in text. All computations yield clean results.\n\n[geometry_trigonometry] (MCQ + 2 SPR)\n- Topics: circle equations, Pythagorean theorem, special right triangles, coordinate geometry, trig ratios, volume/surface area, angle relationships\n- Optional \"visual\" field: {\"type\":\"coordinate_plane\",\"description\":\"...\"}\n- Radical answers OK (e.g., $5\\sqrt{13}$). Multi-step reasoning required.\n\n[statistics_probability] (MCQ + 2 SPR)\n- Topics: mean/median/mode, standard deviation (conceptual), probability (conditional from tables), data interpretation, scatter plots/line of best fit, sampling/bias\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Distractors: P(A|B) vs P(B|A) confusion, mean vs median, causal from correlational\n\n[advanced_math] (MCQ + 2 SPR)\n- Topics: quadratics (factoring, discriminant), polynomials (zeros, end behavior), exponentials (growth/decay), rational expressions, radical expressions, non-linear systems\n- Hardest questions on the test. Multi-step reasoning required. Elegant algebraic insight preferred.\n- All expressions must be factorable with integer coefficients. SPR answers are clean numbers.\n\nVERIFY EVERY ANSWER: Solve each problem step-by-step before finalizing correct_answer.\nOutput ONLY a valid JSON array of EXACTLY 7 objects. No markdown. Count them before outputting.", "text": "```json\n[\n {\n \"id\": \"sat11-math-m1-q1\",\n \"number\": 1,\n \"type\": \"algebra\",\n \"prompt\": \"A taxi service charges a $2.50 pickup fee plus $1.75 per mile. If a passenger's total fare was $20.00, how many miles did the taxi travel?\",\n \"choices\": {\n \"A\": \"8\",\n \"B\": \"10\",\n \"C\": \"11.43\",\n \"D\": \"12.86\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"B\"\n },\n {\n \"id\": \"sat11-math-m1-q2\",\n \"number\": 2,\n \"type\": \"algebra\",\n \"prompt\": \"A small theater sells adult tickets for $12 each and child tickets for $8 each. If a total of 150 tickets were sold for a total of $1,560, how many adult tickets were sold?\",\n \"choices\": {\n \"A\": \"60\",\n \"B\": \"75\",\n \"C\": \"90\",\n \"D\": \"100\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m1-q3\",\n \"number\": 3,\n \"type\": \"algebra\",\n \"prompt\": \"Which of the following inequalities is equivalent to $7 - 2x > 15$?\",\n \"choices\": {\n \"A\": \"$x < -4$\",\n \"B\": \"$x > -4$\",\n \"C\": \"$x < 4$\",\n \"D\": \"$x > 4$\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\"\n },\n {\n \"id\": \"sat11-math-m1-q4\",\n \"number\": 4,\n \"type\": \"algebra\",\n \"prompt\": \"If $|\\frac{x}{2} - 3| = 4$, what is a possible value of $x$?\",\n \"choices\": {\n \"A\": \"2\",\n \"B\": \"7\",\n \"C\": \"-1\",\n \"D\": \"14\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"D\"\n },\n {\n \"id\": \"sat11-math-m1-q5\",\n \"number\": 5,\n \"type\": \"algebra\",\n \"prompt\": \"If $3x + 2y = 18$ and $2x + 3y = 17$, what is the value of $x+y$?\",\n \"choices\": {\n \"A\": \"1\",\n \"B\": \"4\",\n \"C\": \"7\",\n \"D\": \"35\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m1-q6\",\n \"number\": 6,\n \"type\": \"algebra\",\n \"prompt\": \"A rental company charges $40 for the first day and $25 for each additional day to rent a machine. If a rental bill for a machine was $215, for how many total days was it rented?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": \"8\"\n },\n {\n \"id\": \"sat11-math-m1-q7\",\n \"number\": 7,\n \"type\": \"algebra\",\n \"prompt\": \"The inequality $|n - 7| \\le 2$ describes the range of possible values for a variable $n$. If $n$ must be an integer, what is the largest possible value of $n$?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": \"9\"\n }\n]\n```" }, "status": "ok" }, { "attempts": 1, "index": 1, "output": { "costUsd": 0, "model": "gemini-2.5-pro", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Math questions (of which 1 are student-produced-response and the rest are MCQ).\nType: functions_analysis | Test number: 11 | Module: 1 | First question number: 8\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-math-m1-qN\" where N is the sequential question number starting at 8.\nFor example, the first question in this batch has id \"sat11-math-m1-q8\".\n\nMCQ Schema: {\"id\":\"sat11-math-m1-qN\",\"number\":N,\"type\":\"functions_analysis\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\nSPR Schema: {\"id\":\"sat11-math-m1-qN\",\"number\":N,\"type\":\"functions_analysis\",\"prompt\":\"...\",\"answer_format\":\"student_produced\",\"correct_answer\":\"numeric_value\"}\nPlace SPR questions at the END of the array.\n\nFollow instructions for your type:\n\n[algebra] (MCQ + 1 SPR)\n- Topics: linear equations (1 var, 2 vars), systems of 2 linear equations, linear inequalities, absolute value\n- Mix word problems and abstract algebra. Use $...$ notation.\n- MCQ distractors: sign errors, arithmetic mistakes, setup errors, partial answers\n- SPR answers: clean integers or simple fractions\n\n[functions_analysis] (MCQ + 1 SPR)\n- Topics: function evaluation, domain/range, transformations (f(x)+k, f(x-h)), graphical analysis, composition f(g(x)), piecewise\n- At least one conceptual insight question (e.g., (p,q) and (q,p) on same graph)\n- Describe graphs in text. All computations yield clean results.\n\n[geometry_trigonometry] (MCQ + 1 SPR)\n- Topics: circle equations, Pythagorean theorem, special right triangles, coordinate geometry, trig ratios, volume/surface area, angle relationships\n- Optional \"visual\" field: {\"type\":\"coordinate_plane\",\"description\":\"...\"}\n- Radical answers OK (e.g., $5\\sqrt{13}$). Multi-step reasoning required.\n\n[statistics_probability] (MCQ + 1 SPR)\n- Topics: mean/median/mode, standard deviation (conceptual), probability (conditional from tables), data interpretation, scatter plots/line of best fit, sampling/bias\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Distractors: P(A|B) vs P(B|A) confusion, mean vs median, causal from correlational\n\n[advanced_math] (MCQ + 1 SPR)\n- Topics: quadratics (factoring, discriminant), polynomials (zeros, end behavior), exponentials (growth/decay), rational expressions, radical expressions, non-linear systems\n- Hardest questions on the test. Multi-step reasoning required. Elegant algebraic insight preferred.\n- All expressions must be factorable with integer coefficients. SPR answers are clean numbers.\n\nVERIFY EVERY ANSWER: Solve each problem step-by-step before finalizing correct_answer.\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "```json\n[\n {\n \"id\": \"sat11-math-m1-q8\",\n \"number\": 8,\n \"type\": \"functions_analysis\",\n \"prompt\": \"For the function $f$ defined by $f(x) = 3x^2 - 5x + 1$, what is the value of $f(-2)$?\",\n \"choices\": {\n \"A\": \"3\",\n \"B\": \"-1\",\n \"C\": \"23\",\n \"D\": \"-21\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m1-q9\",\n \"number\": 9,\n \"type\": \"functions_analysis\",\n \"prompt\": \"The graph of the function $g$ in the $xy$-plane contains the point $(2, 5)$. Which of the following points must be on the graph of the function $h$ defined by $h(x) = g(x-3) + 4$?\",\n \"choices\": {\n \"A\": \"(-1, 9)\",\n \"B\": \"(5, 9)\",\n \"C\": \"(5, 1)\",\n \"D\": \"(-1, 1)\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"B\"\n },\n {\n \"id\": \"sat11-math-m1-q10\",\n \"number\": 10,\n \"type\": \"functions_analysis\",\n \"prompt\": \"Let the functions $f$ and $g$ be defined by $f(x) = x^2 + 4$ and $g(x) = \\\\sqrt{x-1}$. What is the domain of the composite function $h(x) = f(g(x))$?\",\n \"choices\": {\n \"A\": \"x \\\\ge 1\",\n \"B\": \"x \\\\ge -4\",\n \"C\": \"x \\\\le 1\",\n \"D\": \"All real numbers\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\"\n },\n {\n \"id\": \"sat11-math-m1-q11\",\n \"number\": 11,\n \"type\": \"functions_analysis\",\n \"prompt\": \"A function $f$ is defined as follows:\\n$f(x) = \\\\begin{cases} 2x + 7, & \\\\text{if } x < 3 \\\\\\\\ x^2 - 1, & \\\\text{if } x \\\\ge 3 \\\\end{cases}$\\nWhat is the value of $f(3)$?\",\n \"choices\": {\n \"A\": \"13\",\n \"B\": \"10\",\n \"C\": \"9\",\n \"D\": \"8\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"D\"\n },\n {\n \"id\": \"sat11-math-m1-q12\",\n \"number\": 12,\n \"type\": \"functions_analysis\",\n \"prompt\": \"The functions $g$ and $h$ are defined by $g(x) = 2x - 5$ and $h(x) = x^2 + 3$. What is the value of $g(h(2))$?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": \"9\"\n }\n]\n```" }, "status": "ok" }, { "attempts": 1, "index": 2, "output": { "costUsd": 0, "model": "gemini-2.5-pro", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Math questions (of which 1 are student-produced-response and the rest are MCQ).\nType: geometry_trigonometry | Test number: 11 | Module: 1 | First question number: 13\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-math-m1-qN\" where N is the sequential question number starting at 13.\nFor example, the first question in this batch has id \"sat11-math-m1-q13\".\n\nMCQ Schema: {\"id\":\"sat11-math-m1-qN\",\"number\":N,\"type\":\"geometry_trigonometry\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\nSPR Schema: {\"id\":\"sat11-math-m1-qN\",\"number\":N,\"type\":\"geometry_trigonometry\",\"prompt\":\"...\",\"answer_format\":\"student_produced\",\"correct_answer\":\"numeric_value\"}\nPlace SPR questions at the END of the array.\n\nFollow instructions for your type:\n\n[algebra] (MCQ + 1 SPR)\n- Topics: linear equations (1 var, 2 vars), systems of 2 linear equations, linear inequalities, absolute value\n- Mix word problems and abstract algebra. Use $...$ notation.\n- MCQ distractors: sign errors, arithmetic mistakes, setup errors, partial answers\n- SPR answers: clean integers or simple fractions\n\n[functions_analysis] (MCQ + 1 SPR)\n- Topics: function evaluation, domain/range, transformations (f(x)+k, f(x-h)), graphical analysis, composition f(g(x)), piecewise\n- At least one conceptual insight question (e.g., (p,q) and (q,p) on same graph)\n- Describe graphs in text. All computations yield clean results.\n\n[geometry_trigonometry] (MCQ + 1 SPR)\n- Topics: circle equations, Pythagorean theorem, special right triangles, coordinate geometry, trig ratios, volume/surface area, angle relationships\n- Optional \"visual\" field: {\"type\":\"coordinate_plane\",\"description\":\"...\"}\n- Radical answers OK (e.g., $5\\sqrt{13}$). Multi-step reasoning required.\n\n[statistics_probability] (MCQ + 1 SPR)\n- Topics: mean/median/mode, standard deviation (conceptual), probability (conditional from tables), data interpretation, scatter plots/line of best fit, sampling/bias\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Distractors: P(A|B) vs P(B|A) confusion, mean vs median, causal from correlational\n\n[advanced_math] (MCQ + 1 SPR)\n- Topics: quadratics (factoring, discriminant), polynomials (zeros, end behavior), exponentials (growth/decay), rational expressions, radical expressions, non-linear systems\n- Hardest questions on the test. Multi-step reasoning required. Elegant algebraic insight preferred.\n- All expressions must be factorable with integer coefficients. SPR answers are clean numbers.\n\nVERIFY EVERY ANSWER: Solve each problem step-by-step before finalizing correct_answer.\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "```json\n[\n {\n \"id\": \"sat11-math-m1-q13\",\n \"number\": 13,\n \"type\": \"geometry_trigonometry\",\n \"prompt\": \"In the xy-plane, a circle has a center at $(3, -5)$ and a radius of 7. Which of the following is the equation of the circle?\",\n \"choices\": {\n \"A\": \"$(x+3)^2 + (y-5)^2 = 7$\",\n \"B\": \"$(x-3)^2 + (y+5)^2 = 49$\",\n \"C\": \"$(x-3)^2 + (y+5)^2 = 7$\",\n \"D\": \"$(x+3)^2 + (y-5)^2 = 49$\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"B\"\n },\n {\n \"id\": \"sat11-math-m1-q14\",\n \"number\": 14,\n \"type\": \"geometry_trigonometry\",\n \"prompt\": \"In right triangle ABC, the right angle is at vertex C. If the length of side AC is 8 and the length of side BC is 15, what is the value of $\\\\cos(A)$?\",\n \"choices\": {\n \"A\": \"$8/17$\",\n \"B\": \"$15/17$\",\n \"C\": \"$8/15$\",\n \"D\": \"$17/8$\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\"\n },\n {\n \"id\": \"sat11-math-m1-q15\",\n \"number\": 15,\n \"type\": \"geometry_trigonometry\",\n \"prompt\": \"Two parallel lines are intersected by a transversal line. The measure of one of the resulting interior angles is $(3x + 10)^\\\\circ$ and the measure of its corresponding consecutive interior angle is $(5x + 30)^\\\\circ$. What is the value of $x$?\",\n \"choices\": {\n \"A\": \"10\",\n \"B\": \"17.5\",\n \"C\": \"20\",\n \"D\": \"40\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m1-q16\",\n \"number\": 16,\n \"type\": \"geometry_trigonometry\",\n \"prompt\": \"In right triangle $XYZ$, the measure of angle $Y$ is $90^\\\\circ$ and the measure of angle $X$ is $60^\\\\circ$. If side $XY$ (the side adjacent to angle $X$) has a length of 5, what is the length of the hypotenuse, $XZ$?\",\n \"choices\": {\n \"A\": \"$5\\\\sqrt{2}$\",\n \"B\": \"$5\\\\sqrt{3}$\",\n \"C\": \"$10\\\\sqrt{3}/3$\",\n \"D\": \"10\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"D\"\n },\n {\n \"id\": \"sat11-math-m1-q17\",\n \"number\": 17,\n \"type\": \"geometry_trigonometry\",\n \"prompt\": \"A rectangular prism has a height of 4 cm, a width of 3 cm, and a length of 12 cm. What is the length, in cm, of the space diagonal that connects two opposite vertices of the prism?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": \"13\"\n }\n]\n```" }, "status": "ok" }, { "attempts": 1, "index": 3, "output": { "costUsd": 0, "model": "gemini-2.5-pro", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Math questions (of which 1 are student-produced-response and the rest are MCQ).\nType: statistics_probability | Test number: 11 | Module: 1 | First question number: 18\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-math-m1-qN\" where N is the sequential question number starting at 18.\nFor example, the first question in this batch has id \"sat11-math-m1-q18\".\n\nMCQ Schema: {\"id\":\"sat11-math-m1-qN\",\"number\":N,\"type\":\"statistics_probability\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\nSPR Schema: {\"id\":\"sat11-math-m1-qN\",\"number\":N,\"type\":\"statistics_probability\",\"prompt\":\"...\",\"answer_format\":\"student_produced\",\"correct_answer\":\"numeric_value\"}\nPlace SPR questions at the END of the array.\n\nFollow instructions for your type:\n\n[algebra] (MCQ + 1 SPR)\n- Topics: linear equations (1 var, 2 vars), systems of 2 linear equations, linear inequalities, absolute value\n- Mix word problems and abstract algebra. Use $...$ notation.\n- MCQ distractors: sign errors, arithmetic mistakes, setup errors, partial answers\n- SPR answers: clean integers or simple fractions\n\n[functions_analysis] (MCQ + 1 SPR)\n- Topics: function evaluation, domain/range, transformations (f(x)+k, f(x-h)), graphical analysis, composition f(g(x)), piecewise\n- At least one conceptual insight question (e.g., (p,q) and (q,p) on same graph)\n- Describe graphs in text. All computations yield clean results.\n\n[geometry_trigonometry] (MCQ + 1 SPR)\n- Topics: circle equations, Pythagorean theorem, special right triangles, coordinate geometry, trig ratios, volume/surface area, angle relationships\n- Optional \"visual\" field: {\"type\":\"coordinate_plane\",\"description\":\"...\"}\n- Radical answers OK (e.g., $5\\sqrt{13}$). Multi-step reasoning required.\n\n[statistics_probability] (MCQ + 1 SPR)\n- Topics: mean/median/mode, standard deviation (conceptual), probability (conditional from tables), data interpretation, scatter plots/line of best fit, sampling/bias\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Distractors: P(A|B) vs P(B|A) confusion, mean vs median, causal from correlational\n\n[advanced_math] (MCQ + 1 SPR)\n- Topics: quadratics (factoring, discriminant), polynomials (zeros, end behavior), exponentials (growth/decay), rational expressions, radical expressions, non-linear systems\n- Hardest questions on the test. Multi-step reasoning required. Elegant algebraic insight preferred.\n- All expressions must be factorable with integer coefficients. SPR answers are clean numbers.\n\nVERIFY EVERY ANSWER: Solve each problem step-by-step before finalizing correct_answer.\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "```json\n[\n {\n \"id\": \"sat11-math-m1-q18\",\n \"number\": 18,\n \"type\": \"statistics_probability\",\n \"prompt\": \"A community garden manager models the daily water usage for a plot of tomato plants. The line of best fit for the model is $y = 0.25x + 1.5$, where $y$ is the amount of water used in liters and $x$ is the average daily temperature in degrees Celsius. What is the best interpretation of the slope of this line?\",\n \"choices\": {\n \"A\": \"The initial amount of water used, regardless of temperature, is 0.25 liters.\",\n \"B\": \"For every 1-degree Celsius increase in temperature, the plants require an additional 1.5 liters of water.\",\n \"C\": \"For every 1-degree Celsius increase in average daily temperature, the predicted water usage increases by 0.25 liters.\",\n \"D\": \"For every 0.25-degree Celsius increase in temperature, the water usage increases by 1 liter.\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m1-q19\",\n \"number\": 19,\n \"type\": \"statistics_probability\",\n \"visual\": {\n \"type\": \"table\",\n \"title\": \"Instrument Played by Grade Level\",\n \"headers\": [\"\", \"Strings\", \"Woodwind\", \"Total\"],\n \"rows\": [\n [\"Grade 9\", 18, 22, 40],\n [\"Grade 10\", 24, 16, 40],\n [\"Total\", 42, 38, 80]\n ]\n },\n \"prompt\": \"The table above shows the distribution of 80 students in a music program by grade level and the type of instrument they play. If a student who plays a strings instrument is selected at random, what is the probability that the student is in Grade 9?\",\n \"choices\": {\n \"A\": \"$\\\\frac{18}{80}$\",\n \"B\": \"$\\\\frac{18}{40}$\",\n \"C\": \"$\\\\frac{40}{80}$\",\n \"D\": \"$\\\\frac{18}{42}$\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"D\"\n },\n {\n \"id\": \"sat11-math-m1-q20\",\n \"number\": 20,\n \"type\": \"statistics_probability\",\n \"prompt\": \"The list shows the number of points scored by a basketball player in each of the last 7 games: 15, 21, 18, 24, 19, 35, 17. Which statement accurately compares the mean and median of these scores?\",\n \"choices\": {\n \"A\": \"The mean is greater than the median.\",\n \"B\": \"The median is greater than the mean.\",\n \"C\": \"The mean and the median are equal.\",\n \"D\": \"The relationship cannot be determined from the information given.\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\"\n },\n {\n \"id\": \"sat11-math-m1-q21\",\n \"number\": 21,\n \"type\": \"statistics_probability\",\n \"visual\": {\n \"type\": \"table\",\n \"title\": \"City Employee Departments\",\n \"headers\": [\"Department\", \"Number of Employees\"],\n \"rows\": [\n [\"Public Works\", 450],\n [\"Parks & Recreation\", 200],\n [\"Administration\", 150],\n [\"Safety\", 700]\n ]\n },\n \"prompt\": \"A researcher wants to survey employees about workplace satisfaction in the city described in the table. Which sampling method is most likely to introduce bias?\",\n \"choices\": {\n \"A\": \"Using a random number generator to select 100 employees from a full alphabetical list of all city employees.\",\n \"B\": \"Leaving survey forms at the entrance of the Safety department building and using the first 100 that are returned.\",\n \"C\": \"Randomly selecting 50 employees from the Public Works department.\",\n \"D\": \"Creating a stratified sample by selecting employees from each department in proportion to the department's size.\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"B\"\n },\n {\n \"id\": \"sat11-math-m1-q22\",\n \"number\": 22,\n \"type\": \"statistics_probability\",\n \"prompt\": \"A dataset of 14 numbers has a mean of 68. A fifteenth number is added to the dataset, and the new mean is 70. What is the value of the number that was added to the dataset?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": 98\n }\n]\n```" }, "status": "ok" }, { "attempts": 1, "index": 4, "output": { "costUsd": 0, "model": "gemini-2.5-pro", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Math questions (of which 2 are student-produced-response and the rest are MCQ).\nType: advanced_math | Test number: 11 | Module: 1 | First question number: 23\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-math-m1-qN\" where N is the sequential question number starting at 23.\nFor example, the first question in this batch has id \"sat11-math-m1-q23\".\n\nMCQ Schema: {\"id\":\"sat11-math-m1-qN\",\"number\":N,\"type\":\"advanced_math\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\nSPR Schema: {\"id\":\"sat11-math-m1-qN\",\"number\":N,\"type\":\"advanced_math\",\"prompt\":\"...\",\"answer_format\":\"student_produced\",\"correct_answer\":\"numeric_value\"}\nPlace SPR questions at the END of the array.\n\nFollow instructions for your type:\n\n[algebra] (MCQ + 2 SPR)\n- Topics: linear equations (1 var, 2 vars), systems of 2 linear equations, linear inequalities, absolute value\n- Mix word problems and abstract algebra. Use $...$ notation.\n- MCQ distractors: sign errors, arithmetic mistakes, setup errors, partial answers\n- SPR answers: clean integers or simple fractions\n\n[functions_analysis] (MCQ + 2 SPR)\n- Topics: function evaluation, domain/range, transformations (f(x)+k, f(x-h)), graphical analysis, composition f(g(x)), piecewise\n- At least one conceptual insight question (e.g., (p,q) and (q,p) on same graph)\n- Describe graphs in text. All computations yield clean results.\n\n[geometry_trigonometry] (MCQ + 2 SPR)\n- Topics: circle equations, Pythagorean theorem, special right triangles, coordinate geometry, trig ratios, volume/surface area, angle relationships\n- Optional \"visual\" field: {\"type\":\"coordinate_plane\",\"description\":\"...\"}\n- Radical answers OK (e.g., $5\\sqrt{13}$). Multi-step reasoning required.\n\n[statistics_probability] (MCQ + 2 SPR)\n- Topics: mean/median/mode, standard deviation (conceptual), probability (conditional from tables), data interpretation, scatter plots/line of best fit, sampling/bias\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Distractors: P(A|B) vs P(B|A) confusion, mean vs median, causal from correlational\n\n[advanced_math] (MCQ + 2 SPR)\n- Topics: quadratics (factoring, discriminant), polynomials (zeros, end behavior), exponentials (growth/decay), rational expressions, radical expressions, non-linear systems\n- Hardest questions on the test. Multi-step reasoning required. Elegant algebraic insight preferred.\n- All expressions must be factorable with integer coefficients. SPR answers are clean numbers.\n\nVERIFY EVERY ANSWER: Solve each problem step-by-step before finalizing correct_answer.\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "```json\n[\n {\n \"id\": \"sat11-math-m1-q23\",\n \"number\": 23,\n \"type\": \"advanced_math\",\n \"prompt\": \"In the xy-plane, a parabola with the equation $y = x^2 + bx + c$ and a line with the equation $y = -x + k$ intersect at exactly two points, $(2, 5)$ and $(5, 2)$. What is the value of $b$?\",\n \"choices\": {\n \"A\": \"-8\",\n \"B\": \"-6\",\n \"C\": \"2\",\n \"D\": \"7\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"B\"\n },\n {\n \"id\": \"sat11-math-m1-q24\",\n \"number\": 24,\n \"type\": \"advanced_math\",\n \"prompt\": \"Which of the following expressions is equivalent to $\\\\frac{2x^2-11x+10}{x-4}$ for $x>4$?\",\n \"choices\": {\n \"A\": \"$2x-3 - \\\\frac{2}{x-4}$\",\n \"B\": \"$2x+3 + \\\\frac{22}{x-4}$\",\n \"C\": \"$2x-19 + \\\\frac{86}{x-4}$\",\n \"D\": \"$2x-3 + \\\\frac{22}{x-4}$\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\"\n },\n {\n \"id\": \"sat11-math-m1-q25\",\n \"number\": 25,\n \"type\": \"advanced_math\",\n \"prompt\": \"The function $f$ is defined by $f(x) = k(x+a)(x+b)$, where $k$, $a$, and $b$ are constants. The graph of $y=f(x)$ in the xy-plane is a parabola with x-intercepts at $-3$ and $5$. The vertex of the parabola is $(1, -32)$. What is the value of $k$?\",\n \"choices\": {\n \"A\": \"-2\",\n \"B\": \"1\",\n \"C\": \"2\",\n \"D\": \"8\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m1-q26\",\n \"number\": 26,\n \"type\": \"advanced_math\",\n \"prompt\": \"If $x > 0$ and $\\\\sqrt{3\\\\sqrt{x}} = \\\\sqrt[n]{3x}$ for all values of $x$, what is the value of $n$?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": 4\n },\n {\n \"id\": \"sat11-math-m1-q27\",\n \"number\": 27,\n \"type\": \"advanced_math\",\n \"prompt\": \"For the quadratic equation $kx^2 - 8x + 2 = 0$, the sum of the solutions is twice the product of the solutions. What is the value of $k$?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": 2\n }\n]\n```" }, "status": "ok" }, { "attempts": 1, "index": 5, "output": { "costUsd": 0, "model": "gemini-2.5-pro", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 7 SAT Math questions (of which 2 are student-produced-response and the rest are MCQ).\nType: algebra | Test number: 11 | Module: 2 | First question number: 1\n\nMANDATORY: Your JSON array MUST contain exactly 7 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-math-m2-qN\" where N is the sequential question number starting at 1.\nFor example, the first question in this batch has id \"sat11-math-m2-q1\".\n\nMCQ Schema: {\"id\":\"sat11-math-m2-qN\",\"number\":N,\"type\":\"algebra\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\nSPR Schema: {\"id\":\"sat11-math-m2-qN\",\"number\":N,\"type\":\"algebra\",\"prompt\":\"...\",\"answer_format\":\"student_produced\",\"correct_answer\":\"numeric_value\"}\nPlace SPR questions at the END of the array.\n\nFollow instructions for your type:\n\n[algebra] (MCQ + 2 SPR)\n- Topics: linear equations (1 var, 2 vars), systems of 2 linear equations, linear inequalities, absolute value\n- Mix word problems and abstract algebra. Use $...$ notation.\n- MCQ distractors: sign errors, arithmetic mistakes, setup errors, partial answers\n- SPR answers: clean integers or simple fractions\n\n[functions_analysis] (MCQ + 2 SPR)\n- Topics: function evaluation, domain/range, transformations (f(x)+k, f(x-h)), graphical analysis, composition f(g(x)), piecewise\n- At least one conceptual insight question (e.g., (p,q) and (q,p) on same graph)\n- Describe graphs in text. All computations yield clean results.\n\n[geometry_trigonometry] (MCQ + 2 SPR)\n- Topics: circle equations, Pythagorean theorem, special right triangles, coordinate geometry, trig ratios, volume/surface area, angle relationships\n- Optional \"visual\" field: {\"type\":\"coordinate_plane\",\"description\":\"...\"}\n- Radical answers OK (e.g., $5\\sqrt{13}$). Multi-step reasoning required.\n\n[statistics_probability] (MCQ + 2 SPR)\n- Topics: mean/median/mode, standard deviation (conceptual), probability (conditional from tables), data interpretation, scatter plots/line of best fit, sampling/bias\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Distractors: P(A|B) vs P(B|A) confusion, mean vs median, causal from correlational\n\n[advanced_math] (MCQ + 2 SPR)\n- Topics: quadratics (factoring, discriminant), polynomials (zeros, end behavior), exponentials (growth/decay), rational expressions, radical expressions, non-linear systems\n- Hardest questions on the test. Multi-step reasoning required. Elegant algebraic insight preferred.\n- All expressions must be factorable with integer coefficients. SPR answers are clean numbers.\n\nVERIFY EVERY ANSWER: Solve each problem step-by-step before finalizing correct_answer.\nOutput ONLY a valid JSON array of EXACTLY 7 objects. No markdown. Count them before outputting.", "text": "```json\n[\n {\n \"id\": \"sat11-math-m2-q1\",\n \"number\": 1,\n \"type\": \"algebra\",\n \"prompt\": \"If $\\\\frac{1}{3}(6x + 15) - 4 = 9$, what is the value of $x$?\",\n \"choices\": {\n \"A\": \"-1\",\n \"B\": \"4\",\n \"C\": \"5\",\n \"D\": \"8\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"B\"\n },\n {\n \"id\": \"sat11-math-m2-q2\",\n \"number\": 2,\n \"type\": \"algebra\",\n \"prompt\": \"A gym offers two membership plans. Plan A has a one-time initiation fee of $50 and a monthly fee of $20. Plan B has no initiation fee and a monthly fee of $30. After how many months will the total cost of both plans be equal?\",\n \"choices\": {\n \"A\": \"1.5\",\n \"B\": \"2.5\",\n \"C\": \"5\",\n \"D\": \"10\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m2-q3\",\n \"number\": 3,\n \"type\": \"algebra\",\n \"prompt\": \"Which of the following is a possible value of $y$ if $-4(y - 2) > 12$?\",\n \"choices\": {\n \"A\": \"-2\",\n \"B\": \"-1\",\n \"C\": \"0\",\n \"D\": \"2\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\"\n },\n {\n \"id\": \"sat11-math-m2-q4\",\n \"number\": 4,\n \"type\": \"algebra\",\n \"prompt\": \"If $|2k - 3| = 7$, what is the sum of the two possible values for $k$?\",\n \"choices\": {\n \"A\": \"-2\",\n \"B\": \"0\",\n \"C\": \"3\",\n \"D\": \"5\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m2-q5\",\n \"number\": 5,\n \"type\": \"algebra\",\n \"prompt\": \"If $5x - 2y = 11$ and $x + 3y = 9$, what is the value of $x$?\",\n \"choices\": {\n \"A\": \"-1\",\n \"B\": \"2\",\n \"C\": \"2.5\",\n \"D\": \"3\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"D\"\n },\n {\n \"id\": \"sat11-math-m2-q6\",\n \"number\": 6,\n \"type\": \"algebra\",\n \"prompt\": \"A rental car company charges a flat daily fee plus a fee per mile driven. A customer who drove 100 miles was charged $80. Another customer drove 150 miles on the same day and was charged $95. What is the company's flat daily fee in dollars?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": 50\n },\n {\n \"id\": \"sat11-math-m2-q7\",\n \"number\": 7,\n \"type\": \"algebra\",\n \"prompt\": \"The inequality $|x - 5| \\\\leq 2$ defines a range of values for $x$. If $x$ must be an integer, what is the sum of all possible values of $x$?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": 25\n }\n]\n```" }, "status": "ok" }, { "attempts": 1, "index": 6, "output": { "costUsd": 0, "model": "gemini-2.5-pro", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Math questions (of which 1 are student-produced-response and the rest are MCQ).\nType: functions_analysis | Test number: 11 | Module: 2 | First question number: 8\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-math-m2-qN\" where N is the sequential question number starting at 8.\nFor example, the first question in this batch has id \"sat11-math-m2-q8\".\n\nMCQ Schema: {\"id\":\"sat11-math-m2-qN\",\"number\":N,\"type\":\"functions_analysis\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\nSPR Schema: {\"id\":\"sat11-math-m2-qN\",\"number\":N,\"type\":\"functions_analysis\",\"prompt\":\"...\",\"answer_format\":\"student_produced\",\"correct_answer\":\"numeric_value\"}\nPlace SPR questions at the END of the array.\n\nFollow instructions for your type:\n\n[algebra] (MCQ + 1 SPR)\n- Topics: linear equations (1 var, 2 vars), systems of 2 linear equations, linear inequalities, absolute value\n- Mix word problems and abstract algebra. Use $...$ notation.\n- MCQ distractors: sign errors, arithmetic mistakes, setup errors, partial answers\n- SPR answers: clean integers or simple fractions\n\n[functions_analysis] (MCQ + 1 SPR)\n- Topics: function evaluation, domain/range, transformations (f(x)+k, f(x-h)), graphical analysis, composition f(g(x)), piecewise\n- At least one conceptual insight question (e.g., (p,q) and (q,p) on same graph)\n- Describe graphs in text. All computations yield clean results.\n\n[geometry_trigonometry] (MCQ + 1 SPR)\n- Topics: circle equations, Pythagorean theorem, special right triangles, coordinate geometry, trig ratios, volume/surface area, angle relationships\n- Optional \"visual\" field: {\"type\":\"coordinate_plane\",\"description\":\"...\"}\n- Radical answers OK (e.g., $5\\sqrt{13}$). Multi-step reasoning required.\n\n[statistics_probability] (MCQ + 1 SPR)\n- Topics: mean/median/mode, standard deviation (conceptual), probability (conditional from tables), data interpretation, scatter plots/line of best fit, sampling/bias\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Distractors: P(A|B) vs P(B|A) confusion, mean vs median, causal from correlational\n\n[advanced_math] (MCQ + 1 SPR)\n- Topics: quadratics (factoring, discriminant), polynomials (zeros, end behavior), exponentials (growth/decay), rational expressions, radical expressions, non-linear systems\n- Hardest questions on the test. Multi-step reasoning required. Elegant algebraic insight preferred.\n- All expressions must be factorable with integer coefficients. SPR answers are clean numbers.\n\nVERIFY EVERY ANSWER: Solve each problem step-by-step before finalizing correct_answer.\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "```json\n[\n {\n \"id\": \"sat11-math-m2-q8\",\n \"number\": 8,\n \"type\": \"functions_analysis\",\n \"prompt\": \"Let the function $h$ be defined by $h(x) = x^2 - 3x$. If $h(a+2) = 4$ for some constant $a$, what is one possible value of $a$?\",\n \"choices\": {\n \"A\": \"2\",\n \"B\": \"-1\",\n \"C\": \"0\",\n \"D\": \"4\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\"\n },\n {\n \"id\": \"sat11-math-m2-q9\",\n \"number\": 9,\n \"type\": \"functions_analysis\",\n \"prompt\": \"The graph of a function $f$ in the $xy$-plane passes through the point $(0, 3)$. A second function, $g$, is defined by the transformation $g(x) = f(x-2) + 1$. Which of the following points must be on the graph of $g$?\",\n \"choices\": {\n \"A\": \"(-2, 4)\",\n \"B\": \"(2, 4)\",\n \"C\": \"(2, 2)\",\n \"D\": \"(-2, 2)\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"B\"\n },\n {\n \"id\": \"sat11-math-m2-q10\",\n \"number\": 10,\n \"type\": \"functions_analysis\",\n \"prompt\": \"A function $f$ is defined as $f(x) = \\\\begin{cases} 2x + 7 & \\\\text{if } x \\\\le 1 \\\\\\\\ x^2 - 4 & \\\\text{if } x > 1 \\\\end{cases}$. What is the value of $f(-3) + f(3)$?\",\n \"choices\": {\n \"A\": \"14\",\n \"B\": \"10\",\n \"C\": \"6\",\n \"D\": \"4\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m2-q11\",\n \"number\": 11,\n \"type\": \"functions_analysis\",\n \"prompt\": \"The function $g$ is defined by $g(x) = \\\\frac{x+3}{\\\\sqrt{4-x}}$. For the function's output to be a real number, which of the following inequalities describes all possible values of the input $x$?\",\n \"choices\": {\n \"A\": \"$x > 4$\",\n \"B\": \"$x \\\\ge 4$\",\n \"C\": \"$x \\\\le 4$\",\n \"D\": \"$x < 4$\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"D\"\n },\n {\n \"id\": \"sat11-math-m2-q12\",\n \"number\": 12,\n \"type\": \"functions_analysis\",\n \"prompt\": \"Let the functions $f$ and $g$ be defined by $f(x) = 2x - 5$ and $g(x) = x^2 + k$, where $k$ is a constant. If $f(g(1)) = -3$, what is the value of $k$?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": \"0\"\n }\n]\n```" }, "status": "ok" }, { "attempts": 1, "index": 7, "output": { "costUsd": 0, "model": "gemini-2.5-pro", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Math questions (of which 1 are student-produced-response and the rest are MCQ).\nType: geometry_trigonometry | Test number: 11 | Module: 2 | First question number: 13\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-math-m2-qN\" where N is the sequential question number starting at 13.\nFor example, the first question in this batch has id \"sat11-math-m2-q13\".\n\nMCQ Schema: {\"id\":\"sat11-math-m2-qN\",\"number\":N,\"type\":\"geometry_trigonometry\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\nSPR Schema: {\"id\":\"sat11-math-m2-qN\",\"number\":N,\"type\":\"geometry_trigonometry\",\"prompt\":\"...\",\"answer_format\":\"student_produced\",\"correct_answer\":\"numeric_value\"}\nPlace SPR questions at the END of the array.\n\nFollow instructions for your type:\n\n[algebra] (MCQ + 1 SPR)\n- Topics: linear equations (1 var, 2 vars), systems of 2 linear equations, linear inequalities, absolute value\n- Mix word problems and abstract algebra. Use $...$ notation.\n- MCQ distractors: sign errors, arithmetic mistakes, setup errors, partial answers\n- SPR answers: clean integers or simple fractions\n\n[functions_analysis] (MCQ + 1 SPR)\n- Topics: function evaluation, domain/range, transformations (f(x)+k, f(x-h)), graphical analysis, composition f(g(x)), piecewise\n- At least one conceptual insight question (e.g., (p,q) and (q,p) on same graph)\n- Describe graphs in text. All computations yield clean results.\n\n[geometry_trigonometry] (MCQ + 1 SPR)\n- Topics: circle equations, Pythagorean theorem, special right triangles, coordinate geometry, trig ratios, volume/surface area, angle relationships\n- Optional \"visual\" field: {\"type\":\"coordinate_plane\",\"description\":\"...\"}\n- Radical answers OK (e.g., $5\\sqrt{13}$). Multi-step reasoning required.\n\n[statistics_probability] (MCQ + 1 SPR)\n- Topics: mean/median/mode, standard deviation (conceptual), probability (conditional from tables), data interpretation, scatter plots/line of best fit, sampling/bias\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Distractors: P(A|B) vs P(B|A) confusion, mean vs median, causal from correlational\n\n[advanced_math] (MCQ + 1 SPR)\n- Topics: quadratics (factoring, discriminant), polynomials (zeros, end behavior), exponentials (growth/decay), rational expressions, radical expressions, non-linear systems\n- Hardest questions on the test. Multi-step reasoning required. Elegant algebraic insight preferred.\n- All expressions must be factorable with integer coefficients. SPR answers are clean numbers.\n\nVERIFY EVERY ANSWER: Solve each problem step-by-step before finalizing correct_answer.\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "```json\n[\n {\n \"id\": \"sat11-math-m2-q13\",\n \"number\": 13,\n \"type\": \"geometry_trigonometry\",\n \"prompt\": \"In the xy-plane, a circle has a diameter with endpoints at $(2, -5)$ and $(10, 1)$. Which of the following is an equation of the circle?\",\n \"choices\": {\n \"A\": \"$(x - 6)^2 + (y + 2)^2 = 100$\",\n \"B\": \"$(x + 6)^2 + (y - 2)^2 = 25$\",\n \"C\": \"$(x - 4)^2 + (y + 2)^2 = 50$\",\n \"D\": \"$(x - 6)^2 + (y + 2)^2 = 25$\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"D\"\n },\n {\n \"id\": \"sat11-math-m2-q14\",\n \"number\": 14,\n \"type\": \"geometry_trigonometry\",\n \"prompt\": \"In right triangle $ABC$, the measure of angle $C$ is $90^\\\\circ$. If the length of side $AB$ is 13 and the length of side $BC$ is 5, what is the value of $\\\\tan(A)$?\",\n \"choices\": {\n \"A\": \"$\\\\frac{5}{13}$\",\n \"B\": \"$\\\\frac{12}{13}$\",\n \"C\": \"$\\\\frac{5}{12}$\",\n \"D\": \"$\\\\frac{12}{5}$\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m2-q15\",\n \"number\": 15,\n \"type\": \"geometry_trigonometry\",\n \"prompt\": \"A right circular cylinder and a right circular cone have the same radius and equal volumes. If the height of the cylinder is 12 inches, what is the height of the cone, in inches?\",\n \"choices\": {\n \"A\": \"36\",\n \"B\": \"12\",\n \"C\": \"4\",\n \"D\": \"24\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\"\n },\n {\n \"id\": \"sat11-math-m2-q16\",\n \"number\": 16,\n \"type\": \"geometry_trigonometry\",\n \"prompt\": \"In the xy-plane, $\\\\triangle PQR$ has vertices $P(0,0)$, $Q(6,0)$, and $R(3,4)$. If $\\\\triangle STU$ is similar to $\\\\triangle PQR$, with vertex $S$ at $(0,0)$ and vertex $T$ at $(9,0)$, what is the y-coordinate of vertex $U$?\",\n \"choices\": {\n \"A\": \"$\\\\frac{8}{3}$\",\n \"B\": \"6\",\n \"C\": \"4\",\n \"D\": \"7\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"B\"\n },\n {\n \"id\": \"sat11-math-m2-q17\",\n \"number\": 17,\n \"type\": \"geometry_trigonometry\",\n \"prompt\": \"A regular hexagon is inscribed in a circle. What is the measure in degrees of the acute angle formed between a side of the hexagon and a line tangent to the circle at a shared vertex?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": 30\n }\n]\n```" }, "status": "ok" }, { "attempts": 1, "index": 8, "output": { "costUsd": 0, "model": "gemini-2.5-pro", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Math questions (of which 1 are student-produced-response and the rest are MCQ).\nType: statistics_probability | Test number: 11 | Module: 2 | First question number: 18\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-math-m2-qN\" where N is the sequential question number starting at 18.\nFor example, the first question in this batch has id \"sat11-math-m2-q18\".\n\nMCQ Schema: {\"id\":\"sat11-math-m2-qN\",\"number\":N,\"type\":\"statistics_probability\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\nSPR Schema: {\"id\":\"sat11-math-m2-qN\",\"number\":N,\"type\":\"statistics_probability\",\"prompt\":\"...\",\"answer_format\":\"student_produced\",\"correct_answer\":\"numeric_value\"}\nPlace SPR questions at the END of the array.\n\nFollow instructions for your type:\n\n[algebra] (MCQ + 1 SPR)\n- Topics: linear equations (1 var, 2 vars), systems of 2 linear equations, linear inequalities, absolute value\n- Mix word problems and abstract algebra. Use $...$ notation.\n- MCQ distractors: sign errors, arithmetic mistakes, setup errors, partial answers\n- SPR answers: clean integers or simple fractions\n\n[functions_analysis] (MCQ + 1 SPR)\n- Topics: function evaluation, domain/range, transformations (f(x)+k, f(x-h)), graphical analysis, composition f(g(x)), piecewise\n- At least one conceptual insight question (e.g., (p,q) and (q,p) on same graph)\n- Describe graphs in text. All computations yield clean results.\n\n[geometry_trigonometry] (MCQ + 1 SPR)\n- Topics: circle equations, Pythagorean theorem, special right triangles, coordinate geometry, trig ratios, volume/surface area, angle relationships\n- Optional \"visual\" field: {\"type\":\"coordinate_plane\",\"description\":\"...\"}\n- Radical answers OK (e.g., $5\\sqrt{13}$). Multi-step reasoning required.\n\n[statistics_probability] (MCQ + 1 SPR)\n- Topics: mean/median/mode, standard deviation (conceptual), probability (conditional from tables), data interpretation, scatter plots/line of best fit, sampling/bias\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Distractors: P(A|B) vs P(B|A) confusion, mean vs median, causal from correlational\n\n[advanced_math] (MCQ + 1 SPR)\n- Topics: quadratics (factoring, discriminant), polynomials (zeros, end behavior), exponentials (growth/decay), rational expressions, radical expressions, non-linear systems\n- Hardest questions on the test. Multi-step reasoning required. Elegant algebraic insight preferred.\n- All expressions must be factorable with integer coefficients. SPR answers are clean numbers.\n\nVERIFY EVERY ANSWER: Solve each problem step-by-step before finalizing correct_answer.\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "```json\n[\n {\n \"id\": \"sat11-math-m2-q18\",\n \"number\": 18,\n \"type\": \"statistics_probability\",\n \"prompt\": \"The scatterplot above shows the relationship between two variables, $x$ and $y$. A line of best fit for the data is also shown. For the data point $(10, 55)$, the actual $y$-value is 55. What is the $y$-value predicted by the line of best fit when $x=10$, and what is the value of the corresponding residual?\",\n \"visual\": {\n \"type\": \"coordinate_plane\",\n \"description\": \"A scatterplot with about 15 points that show a strong, positive, linear association. The line of best fit is drawn through the points. The line passes through approximately (0, 10) and (20, 90). The point (10, 55) is clearly plotted slightly above the line of best fit.\"\n },\n \"choices\": {\n \"A\": \"Predicted value: 50; Residual: -5\",\n \"B\": \"Predicted value: 55; Residual: 0\",\n \"C\": \"Predicted value: 50; Residual: 5\",\n \"D\": \"Predicted value: 5; Residual: 50\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m2-q19\",\n \"number\": 19,\n \"type\": \"statistics_probability\",\n \"prompt\": \"The table above summarizes the results of a survey of 200 smartphone users on their preferred mobile operating system and their primary use for the phone. Based on the table, what is the probability that a user who primarily uses their phone for communication prefers OS Y?\",\n \"visual\": {\n \"type\": \"table\",\n \"title\": \"Smartphone User Survey\",\n \"headers\": [\"Primary Use\", \"OS X\", \"OS Y\", \"Total\"],\n \"rows\": [\n [\"Communication\", 60, 40, 100],\n [\"Entertainment\", 70, 30, 100],\n [\"Total\", 130, 70, 200]\n ]\n },\n \"choices\": {\n \"A\": \"0.20\",\n \"B\": \"0.35\",\n \"C\": \"0.57\",\n \"D\": \"0.40\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"D\"\n },\n {\n \"id\": \"sat11-math-m2-q20\",\n \"number\": 20,\n \"type\": \"statistics_probability\",\n \"prompt\": \"A small business has 11 employees. The annual salaries, in thousands of dollars, are: 45, 45, 48, 50, 52, 55, 58, 60, 65, 70, and 300. Which of the following statements best compares the mean and median of these salaries?\",\n \"choices\": {\n \"A\": \"The mean is greater than the median.\",\n \"B\": \"The median is greater than the mean.\",\n \"C\": \"The mean and the median are equal.\",\n \"D\": \"The relationship between the mean and median cannot be determined.\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\"\n },\n {\n \"id\": \"sat11-math-m2-q21\",\n \"number\": 21,\n \"type\": \"statistics_probability\",\n \"prompt\": \"A market researcher for a national coffee chain wants to determine the average amount of money a customer spends per visit. Which of the following sampling methods is most likely to introduce a significant bias into the study?\",\n \"choices\": {\n \"A\": \"Selecting 100 customers at random from the company's national loyalty program database.\",\n \"B\": \"Surveying every customer who visits a single, specific store location between 8 a.m. and 9 a.m. on a Monday morning.\",\n \"C\": \"Dividing the chain's stores into regions and randomly selecting 5 stores from each region, then surveying all customers at those stores for one day.\",\n \"D\": \"Using a computer to randomly select 200 transaction times from the previous week and surveying the customer who made each of those transactions.\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"B\"\n },\n {\n \"id\": \"sat11-math-m2-q22\",\n \"number\": 22,\n \"type\": \"statistics_probability\",\n \"prompt\": \"At a conference of 300 attendees, 180 are from North America and the rest are from Europe. If 60% of the North American attendees are scientists and 40% of the European attendees are scientists, what is the probability that an attendee selected at random is a scientist? Express your answer as a decimal.\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": \"0.52\"\n }\n]\n```" }, "status": "ok" }, { "attempts": 1, "index": 9, "output": { "costUsd": 0, "model": "gemini-2.5-pro", "prompt": "You are one of the world's foremost standardized test question writers with 25 years of College Board SAT experience. Your questions pass psychometric review on first submission.\n\nOUTPUT FORMAT (mandatory):\nYour entire response MUST be a single JSON array. No prose. No markdown fences. No leading/trailing text.\nEvery MCQ question object has this EXACT structure — only these keys, in this order:\n{\n \"id\": \"string\",\n \"number\": integer,\n \"type\": \"string\",\n \"passage\": \"string (omit for pure-math)\",\n \"prompt\": \"string\",\n \"choices\": {\"A\": \"string\", \"B\": \"string\", \"C\": \"string\", \"D\": \"string\"},\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\" | \"B\" | \"C\" | \"D\"\n}\nThe \"choices\" object has EXACTLY four keys: A, B, C, D. It contains no other keys. Every choice value is a non-empty string.\nFor student-produced-response questions, omit \"choices\" and set \"answer_format\": \"student_produced\" with numeric \"correct_answer\".\n\nABSOLUTE RULES (violation = rejection):\n1. The array contains EXACTLY the requested count of questions — not fewer, not more.\n2. Each question has EXACTLY ONE defensible correct answer (no tie cases).\n3. Every question is 100% original (no recycled phrasing from known tests).\n4. Every distractor represents a real, common student mistake — not nonsense or trivially wrong.\n5. Zero cultural, racial, gender, or socioeconomic bias.\n6. Use $...$ LaTeX for ALL math notation (variables, equations, fractions, exponents).\n7. Reading passages must feel like genuine published academic/literary excerpts.\n8. VERIFY your own answer key before finalizing — mentally solve/evaluate every question.\n\nDISTRACTOR QUALITY:\n- Vocabulary: near-synonyms from the same semantic field that fail in THIS specific context\n- Reading: plausible misreadings (wrong scope, surface-level, opposite, over-generalization)\n- Math MCQ: each wrong answer = a specific common error (sign flip, wrong formula, off-by-one, partial simplification)\n- Grammar: each wrong choice introduces ONE specific grammatical error\n- At least 2 distractors should require careful analysis to eliminate.\n\nANSWER DISTRIBUTION: Distribute correct answers evenly across A, B, C, D.\n- In a batch of N questions, each letter should appear roughly N/4 times as correct.\n- Never have 3+ consecutive same-letter answers.\n- Before finalizing, count correct_answer values; if any letter deviates by more than 2 from N/4, rebalance.\n\nSELF-CHECK before output:\n□ Array length equals requested count exactly\n□ Every \"choices\" object has precisely 4 keys: A, B, C, D\n□ No \"answer_format\" or \"correct_answer\" inside \"choices\"\n□ Every correct_answer is one of \"A\", \"B\", \"C\", \"D\" (or numeric for SPR)\n□ Answer letters are balanced across the batch\n\nGenerate EXACTLY 5 SAT Math questions (of which 2 are student-produced-response and the rest are MCQ).\nType: advanced_math | Test number: 11 | Module: 2 | First question number: 23\n\nMANDATORY: Your JSON array MUST contain exactly 5 question objects.\nEach MCQ has exactly 4 choices: A, B, C, D (no E, no empty values).\n\nID format: each question id MUST be \"sat11-math-m2-qN\" where N is the sequential question number starting at 23.\nFor example, the first question in this batch has id \"sat11-math-m2-q23\".\n\nMCQ Schema: {\"id\":\"sat11-math-m2-qN\",\"number\":N,\"type\":\"advanced_math\",\"prompt\":\"...\",\"choices\":{\"A\":\"...\",\"B\":\"...\",\"C\":\"...\",\"D\":\"...\"},\"answer_format\":\"mcq\",\"correct_answer\":\"X\"}\nSPR Schema: {\"id\":\"sat11-math-m2-qN\",\"number\":N,\"type\":\"advanced_math\",\"prompt\":\"...\",\"answer_format\":\"student_produced\",\"correct_answer\":\"numeric_value\"}\nPlace SPR questions at the END of the array.\n\nFollow instructions for your type:\n\n[algebra] (MCQ + 2 SPR)\n- Topics: linear equations (1 var, 2 vars), systems of 2 linear equations, linear inequalities, absolute value\n- Mix word problems and abstract algebra. Use $...$ notation.\n- MCQ distractors: sign errors, arithmetic mistakes, setup errors, partial answers\n- SPR answers: clean integers or simple fractions\n\n[functions_analysis] (MCQ + 2 SPR)\n- Topics: function evaluation, domain/range, transformations (f(x)+k, f(x-h)), graphical analysis, composition f(g(x)), piecewise\n- At least one conceptual insight question (e.g., (p,q) and (q,p) on same graph)\n- Describe graphs in text. All computations yield clean results.\n\n[geometry_trigonometry] (MCQ + 2 SPR)\n- Topics: circle equations, Pythagorean theorem, special right triangles, coordinate geometry, trig ratios, volume/surface area, angle relationships\n- Optional \"visual\" field: {\"type\":\"coordinate_plane\",\"description\":\"...\"}\n- Radical answers OK (e.g., $5\\sqrt{13}$). Multi-step reasoning required.\n\n[statistics_probability] (MCQ + 2 SPR)\n- Topics: mean/median/mode, standard deviation (conceptual), probability (conditional from tables), data interpretation, scatter plots/line of best fit, sampling/bias\n- Include \"visual\" field (table) for at least half: {\"type\":\"table\",\"title\":\"...\",\"headers\":[...],\"rows\":[[...],...]}\n- Distractors: P(A|B) vs P(B|A) confusion, mean vs median, causal from correlational\n\n[advanced_math] (MCQ + 2 SPR)\n- Topics: quadratics (factoring, discriminant), polynomials (zeros, end behavior), exponentials (growth/decay), rational expressions, radical expressions, non-linear systems\n- Hardest questions on the test. Multi-step reasoning required. Elegant algebraic insight preferred.\n- All expressions must be factorable with integer coefficients. SPR answers are clean numbers.\n\nVERIFY EVERY ANSWER: Solve each problem step-by-step before finalizing correct_answer.\nOutput ONLY a valid JSON array of EXACTLY 5 objects. No markdown. Count them before outputting.", "text": "```json\n[\n {\n \"id\": \"sat11-math-m2-q23\",\n \"number\": 23,\n \"type\": \"advanced_math\",\n \"prompt\": \"The parabola with equation $y = x^2 - 5x + 11$ intersects the line with equation $y = 3x - 4$ at two points, $(x_1, y_1)$ and $(x_2, y_2)$. What is the value of $x_1 + x_2$?\",\n \"choices\": {\n \"A\": \"8\",\n \"B\": \"15\",\n \"C\": \"2\",\n \"D\": \"-2\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"A\"\n },\n {\n \"id\": \"sat11-math-m2-q24\",\n \"number\": 24,\n \"type\": \"advanced_math\",\n \"prompt\": \"The expression $\\\\frac{3x^2+5x-2}{x^2-4}$ can be written as $A + \\\\frac{B}{x-2}$, where $A$ and $B$ are constants and $x \\\\neq \\\\pm 2$. What is the value of $B$?\",\n \"choices\": {\n \"A\": \"3\",\n \"B\": \"5\",\n \"C\": \"8.5\",\n \"D\": \"4.5\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"D\"\n },\n {\n \"id\": \"sat11-math-m2-q25\",\n \"number\": 25,\n \"type\": \"advanced_math\",\n \"prompt\": \"For a function defined by $f(x) = c \\cdot p^x$, where $c$ and $p$ are positive constants, $f(1)=15$ and $f(3)=135$. What is the value of $p$?\",\n \"choices\": {\n \"A\": \"9\",\n \"B\": \"5\",\n \"C\": \"3\",\n \"D\": \"60\"\n },\n \"answer_format\": \"mcq\",\n \"correct_answer\": \"C\"\n },\n {\n \"id\": \"sat11-math-m2-q26\",\n \"number\": 26,\n \"type\": \"advanced_math\",\n \"prompt\": \"If $x > 0$ and $2\\sqrt{x+2} = x+1$, what is the value of $x$?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": 2\n },\n {\n \"id\": \"sat11-math-m2-q27\",\n \"number\": 27,\n \"type\": \"advanced_math\",\n \"prompt\": \"In the $xy$-plane, the graphs of the functions $f(x) = -x^2+3$ and $g(x) = x^2+bx-1$ intersect at exactly one point. What is the value of $b$?\",\n \"answer_format\": \"student_produced\",\n \"correct_answer\": 4\n }\n]\n```" }, "status": "ok" } ], "succeeded": 10 } } }

Noeud	Statut	Duree	Tentatives
Assemble Exam	Success	462ms	1
Build Batches	Success	328ms	1
Generate Math (Gemini 3 Pro)	Success	3m 11s	1
Generate RW (GPT 5.4)	Success	2m 41s	1
Test Number	Success	268ms	1
Output	Success	582ms	1

Run run_20260419Success