Datasets:
source_row_id int64 17 446k | source_row_group int64 0 2 | source_row_in_group int64 8 200k | source_dataset stringclasses 1
value | source_revision stringclasses 1
value | source_parquet_sha256 stringclasses 1
value | input_sha256 stringlengths 64 64 | normalized_input_sha256 stringlengths 64 64 | question_stem_sha256 stringlengths 64 64 | contamination_index_sha256 stringclasses 1
value | domain stringclasses 1
value | expected_answer stringclasses 10
values | source_expected_answer stringclasses 10
values | input stringlengths 328 3.78k | output stringlengths 1.07k 116k | verification dict |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
98,971 | 0 | 98,971 | nvidia/OpenScienceReasoning-2 | 174b02c9cdf231f220765b2a1d5ece4550921894 | e82e9c7de7ac4befc12734d1b7b42b62c9896ba5bdef4e5d17cb7e7e8bceaef2 | 48ca1674d2c237ef0ea199f42c67ea8c57deb5c57a06413c400b400e8edf7437 | 37c4ed7dcec4bf8e5d7efd4f67e2363043a1dba06daefc8c9f0f9b46f8ebde6b | 5e3702c9375a31d53561b5c1336062bdcf35e268c3af1399002f1bdf085b2033 | 92d260107f933e0d30ececc18292740a0b25f19492ab32fc14b903000ca97659 | MEDICINE | E | E | Solve the following problem. Make sure to put the answer (and only answer) inside \boxed{}.
In a phase II trial of 16 patients with non-small cell lung cancer, response assessment per RECIST 1.1 showed the following outcomes after treatment: Patient 1: CR at week 8 and week 12 (confirmed CR). Patient 2: Baseline SLD 1... | <think>
First, I need to calculate the investigator-assessed overall response rate (ORR) for this phase II trial with 16 patients. ORR is typically defined as the proportion of patients who achieve a complete response (CR) or partial response (PR) as their best overall response, and it must be confirmed in most cases. ... | {
"domain_classifier": {
"source_row_id": 98971,
"domain_class": "TARGET_DOMAIN",
"observed_domain": "oncology clinical trial response assessment",
"confidence": "high",
"rationale": "Requires RECIST-based oncology response classification and ORR calculation in NSCLC patients.",
"classifier_ad... |
17,529 | 0 | 17,529 | nvidia/OpenScienceReasoning-2 | 174b02c9cdf231f220765b2a1d5ece4550921894 | e82e9c7de7ac4befc12734d1b7b42b62c9896ba5bdef4e5d17cb7e7e8bceaef2 | 9edf52751b013bfc3b06eb604dffc7ad22bfb1345c8b25ca0012d1ea7491b99d | 0d5d48563dfb4d3f24475d7f02861e38612f4911fe4bd49ed42fef62864c1592 | a1fae3bdf80267b9b1a8778a3f359d8c286a682ee0689217fff3f90045cc867c | 92d260107f933e0d30ececc18292740a0b25f19492ab32fc14b903000ca97659 | MEDICINE | C | C | Solve the following problem. Make sure to put the answer (and only answer) inside \boxed{}.
A patient with a severe bacterial infection is prescribed Drug X intravenously. Drug X has a half-life of 8 hours, exhibits first-order kinetics, and has a narrow therapeutic index. It forms an active metabolite Y with a half-l... | <think>
The problem involves calculating the maximum safe maintenance dose for Drug X with changing pharmacokinetics. Let me break it down.
First, Drug X has a half-life of 8 hours and follows first-order kinetics. Its metabolite Y has a half-life of 24 hours. The therapeutic effect requires Drug X concentration above... | {
"domain_classifier": {
"source_row_id": 17529,
"domain_class": "TARGET_DOMAIN",
"observed_domain": "clinical pharmacokinetics / dosing",
"confidence": "high",
"rationale": "Patient dosing for infection with therapeutic and toxicity constraints is clinical pharmacology.",
"classifier_adapter_... |
370,732 | 1 | 170,633 | nvidia/OpenScienceReasoning-2 | 174b02c9cdf231f220765b2a1d5ece4550921894 | e82e9c7de7ac4befc12734d1b7b42b62c9896ba5bdef4e5d17cb7e7e8bceaef2 | dfe2e2675fb52539ae84a81cd8433e69647d5e31e47017931cb082bbb3a8b0c9 | a2dbcad0542b876d6bb34301df499eb87f0b24467b97f8867be28bddbf6f713f | 53176d96b2bb379e9595a897598359cba696d242a0c1aeb5f9e64048aebe8cde | 92d260107f933e0d30ececc18292740a0b25f19492ab32fc14b903000ca97659 | MEDICINE | C | C | Solve the following problem. Make sure to put the answer (and only answer) inside \boxed{}.
In a population, the prevalence of Disease X is 1%, and the prevalence of Disease Y is 2%. The presence of Disease X and Disease Y are not independent: if an individual has Disease X, the probability of having Disease Y is 50%.... | <think>
I need to find the probability that an individual actually has Disease X given that they tested positive for both diseases. The options are A: 84.5%, B: 86.5%, C: 87.0%, D: 89.1%.
First, I am considering the given prevalences and dependencies. The prevalence of X is 1%, so P(X) = 0.01. Prevalence of Y is 2%, s... | {
"domain_classifier": {
"source_row_id": 370732,
"domain_class": "TARGET_DOMAIN",
"observed_domain": "diagnostic testing / Bayesian clinical probability",
"confidence": "high",
"rationale": "Calculates posterior disease probability from test characteristics and comorbidity dependence.",
"clas... |
399,188 | 1 | 199,089 | nvidia/OpenScienceReasoning-2 | 174b02c9cdf231f220765b2a1d5ece4550921894 | e82e9c7de7ac4befc12734d1b7b42b62c9896ba5bdef4e5d17cb7e7e8bceaef2 | 9ad31d7de0e0dc76794d252131c2d7b52a01d2dc93696865095f9ddbf6e62f94 | cb8dbc7bbdd793048f48553c600baca89d200eb61d7d91833cadd06520a0cdb5 | 2fe87837dbbf206937c3dec4b4d6787d1459d9086caa1f5c4084d51a6fd7a5b6 | 92d260107f933e0d30ececc18292740a0b25f19492ab32fc14b903000ca97659 | MEDICINE | B | B | Solve the following problem. Make sure to put the answer (and only answer) inside \boxed{}.
In a clinical trial, a biomarker test is used to assign treatment. The true prevalence of the biomarker is 30%. The test has a sensitivity of 90% and a specificity of 80%. Patients with a positive test result receive Treatment ... | <think>
I need to find the overall probability of a positive response in this clinical trial setup. The biomarker test has a sensitivity of 90% and specificity of 80%, with a true prevalence of 30%. Based on the test result, patients get Treatment A or B, and response rates depend on true biomarker status and treatment... | {
"domain_classifier": {
"source_row_id": 399188,
"domain_class": "TARGET_DOMAIN",
"observed_domain": "biomarker-guided treatment probability",
"confidence": "high",
"rationale": "Clinical treatment assignment based on biomarker test performance and response rates.",
"classifier_adapter_sha256... |
329,921 | 1 | 129,822 | nvidia/OpenScienceReasoning-2 | 174b02c9cdf231f220765b2a1d5ece4550921894 | e82e9c7de7ac4befc12734d1b7b42b62c9896ba5bdef4e5d17cb7e7e8bceaef2 | 8269fd38ea563834f925ac6ee16d93087d06ff578818f1ddca2664dc4836c3a5 | 66a18588cf43fcb4200c71b14843d9b31af3393fba3dee66357af57deaa2bb48 | 04af5abb9c15d170e5faef08498a070cc597d65947fba40a16914f53beaa1bec | 92d260107f933e0d30ececc18292740a0b25f19492ab32fc14b903000ca97659 | MEDICINE | D | D | "Solve the following problem. Make sure to put the answer (and only answer) inside \\boxed{}.\n\nCar(...TRUNCATED) | "<think>\nI need to solve this problem about CardioGuard, a prodrug metabolized by CYP2D6 and CYP3A4(...TRUNCATED) | {"domain_classifier":{"source_row_id":329921,"domain_class":"TARGET_DOMAIN","observed_domain":"pharm(...TRUNCATED) |
100,175 | 0 | 100,175 | nvidia/OpenScienceReasoning-2 | 174b02c9cdf231f220765b2a1d5ece4550921894 | e82e9c7de7ac4befc12734d1b7b42b62c9896ba5bdef4e5d17cb7e7e8bceaef2 | 3cbc327425f5862746e793fa4a66e002fed315563e1aa05f40ecc50c9ccedbb0 | bcfca758f28de178b41bc368d3c0a0d880ec0cb91a69bcd319873e89d973bdd1 | ab38a6085493026c95a7d430cd2ae4f2dc6503fa246856ec332d02c640c8a0f0 | 92d260107f933e0d30ececc18292740a0b25f19492ab32fc14b903000ca97659 | MEDICINE | D | D | "Solve the following problem. Make sure to put the answer (and only answer) inside \\boxed{}.\n\nA p(...TRUNCATED) | "<think>\nWe are given a patient with severe hepatic impairment (Child-Pugh C) and hypoalbuminemia ((...TRUNCATED) | {"domain_classifier":{"source_row_id":100175,"domain_class":"TARGET_DOMAIN","observed_domain":"clini(...TRUNCATED) |
38,641 | 0 | 38,641 | nvidia/OpenScienceReasoning-2 | 174b02c9cdf231f220765b2a1d5ece4550921894 | e82e9c7de7ac4befc12734d1b7b42b62c9896ba5bdef4e5d17cb7e7e8bceaef2 | 1cac993538140b476e8213093342d3e6c21e6809daa77fdb7a667447453b05d9 | d070c23fac36549c96be8d22dd4e1cfb01180a107f3d61d59ce231a54f6490e6 | a9bcf84820c05c6f0b47d684f5838304de57ec6e6fcf5f1e2b234006fe43dc26 | 92d260107f933e0d30ececc18292740a0b25f19492ab32fc14b903000ca97659 | MEDICINE | J | J | "Solve the following problem. Make sure to put the answer (and only answer) inside \\boxed{}.\n\nA p(...TRUNCATED) | "<think>\nFirst, the question is about a patient with oligometastatic HER2-positive breast cancer, s(...TRUNCATED) | {"domain_classifier":{"source_row_id":38641,"domain_class":"TARGET_DOMAIN","observed_domain":"multid(...TRUNCATED) |
110,127 | 0 | 110,127 | nvidia/OpenScienceReasoning-2 | 174b02c9cdf231f220765b2a1d5ece4550921894 | e82e9c7de7ac4befc12734d1b7b42b62c9896ba5bdef4e5d17cb7e7e8bceaef2 | aae28ba967dd3f4279e00e5f579907a21193d2ef55eb36b380cb2a8862a6d4f2 | 4d8b170ac78bad1f409180d4e94f5d97f9142146180390cc42f260adc0ea0647 | 59d802718ea95e65b650e16803ff3f6faa891a84c57a35f9b6cff2c66b3df5a4 | 92d260107f933e0d30ececc18292740a0b25f19492ab32fc14b903000ca97659 | MEDICINE | F | F | "Solve the following problem. Make sure to put the answer (and only answer) inside \\boxed{}.\n\nAcc(...TRUNCATED) | "<think>\nFirst, the question is about identifying which patient with documented atherosclerotic car(...TRUNCATED) | {"domain_classifier":{"source_row_id":110127,"domain_class":"TARGET_DOMAIN","observed_domain":"cardi(...TRUNCATED) |
348,210 | 1 | 148,111 | nvidia/OpenScienceReasoning-2 | 174b02c9cdf231f220765b2a1d5ece4550921894 | e82e9c7de7ac4befc12734d1b7b42b62c9896ba5bdef4e5d17cb7e7e8bceaef2 | 88c4bd6067e6a530b578b8098fbb4775eb185999acb5d541aeb1e3a819fdfe7e | 32c70a92859b221ab567d5860ee3a842a8de17181bdce2344986ac25fbf8eb1a | c4850d3c6b28932ebe1efd26f2fcc251d56058eece51570a6df0d2cd81bcce51 | 92d260107f933e0d30ececc18292740a0b25f19492ab32fc14b903000ca97659 | MEDICINE | B | B | "Solve the following problem. Make sure to put the answer (and only answer) inside \\boxed{}.\n\nA p(...TRUNCATED) | "<think>\nI need to figure out the therapeutic efficacy of Drug X for this patient compared to a ref(...TRUNCATED) | {"domain_classifier":{"source_row_id":348210,"domain_class":"TARGET_DOMAIN","observed_domain":"pharm(...TRUNCATED) |
280,973 | 1 | 80,874 | nvidia/OpenScienceReasoning-2 | 174b02c9cdf231f220765b2a1d5ece4550921894 | e82e9c7de7ac4befc12734d1b7b42b62c9896ba5bdef4e5d17cb7e7e8bceaef2 | c793f11c34f5292b43a6c929ab52388c67ce4ab742d91dc0799af6faccb07313 | 92a5aa7c8661f2c128ab2a0ecd913387824e16c428d2af8e416736ed534b6ee0 | 50ebb3698dfeddff629c43d02f9571f93e94fb46bc3542f0897677c13f328546 | 92d260107f933e0d30ececc18292740a0b25f19492ab32fc14b903000ca97659 | MEDICINE | B | B | "Solve the following problem. Make sure to put the answer (and only answer) inside \\boxed{}.\n\nRes(...TRUNCATED) | "<think>\nI need to determine for which patient the intervention of timed daily oral melatonin and b(...TRUNCATED) | {"domain_classifier":{"source_row_id":280973,"domain_class":"TARGET_DOMAIN","observed_domain":"sleep(...TRUNCATED) |
End of preview. Expand in Data Studio
OpenSciReasoning-Medicine-20K
This is a 20,000-row Medicine subset derived from
nvidia/OpenScienceReasoning-2 for domain-specific reasoner training and
cross-domain transfer experiments.
Each record preserves the stable source_row_id, defined as the zero-based row
index in the frozen source Parquet file.
Acceptance gates
- High-confidence target-domain classification by the strict domain-only classifier.
- Original source
expected_answerretained. - Final answer parsed from the source reasoning trace matches the retained expected answer.
- Unique
source_row_id. - Unique normalized input.
- Benchmark contamination check performed against the pinned local GPQA, SuperGPQA, and MMLU-Pro contamination index.
- Source replay from the frozen OpenScienceReasoning-2 Parquet completed.
Provenance
- Source dataset:
nvidia/OpenScienceReasoning-2 - Source revision:
174b02c9cdf231f220765b2a1d5ece4550921894 - Source Parquet SHA-256:
e82e9c7de7ac4befc12734d1b7b42b62c9896ba5bdef4e5d17cb7e7e8bceaef2 - Data SHA-256:
91e557fed69704cef6255166b19c62473301ec7b3de05e8e99c9c3efffde80b4
Files
medicine_20000.jsonl: accepted training split, 20,000 rows.audit/release_audit.json: final local audit.audit/classification_collect_summary.json: classification collection summary, when available.
Automated filtering substantially improves data quality, but users should still perform task-specific validation before high-stakes use.
- Downloads last month
- 33