Dataset Viewer
Auto-converted to Parquet Duplicate
test_info
dict
summary
dict
overall
dict
detailed_results
listlengths
95
116
{ "date": "0723_163058", "model": "gpt-5.4_2026-03-05", "month": "202603", "reasoning_effort": "medium", "min_score_filter": null, "total_candidates": 112, "total_tested": 112, "timestamp": "2026-07-23T16:51:24", "input_dir": "/mnt/shared/hanze/local/icml2026-research-level-math/Mathematician/helpers/...
{ "ge5_hard": { "correct": 42, "total": 112, "accuracy": 0.375 } }
{ "correct": 42, "total": 112, "accuracy": 0.375 }
[ { "id": "2603.27196v1::orig", "theorem_type": "ge5_hard", "score": 5, "correct_answer": "E", "model_answer": "B", "is_correct": false, "reasoning_effort": "medium", "requested_reasoning_effort": "medium", "instance": "msra/shared", "elapsed_seconds": 15.073241233825684, "...
{ "date": "0723_191846", "model": "gpt-5.4_2026-03-05", "month": "202604", "reasoning_effort": "medium", "min_score_filter": null, "total_candidates": 108, "total_tested": 108, "timestamp": "2026-07-23T19:36:37", "input_dir": "/mnt/shared/hanze/local/icml2026-research-level-math/Mathematician/helpers/...
{ "ge5_hard": { "correct": 46, "total": 108, "accuracy": 0.42592592592592593 } }
{ "correct": 46, "total": 108, "accuracy": 0.42592592592592593 }
[ { "id": "2604.25513v1::alt1", "theorem_type": "ge5_hard", "score": 6, "correct_answer": "E", "model_answer": "E", "is_correct": true, "reasoning_effort": "medium", "requested_reasoning_effort": "medium", "instance": "gcr/shared", "elapsed_seconds": 10.692768573760986, "er...
{ "date": "0723_225357", "model": "gpt-5.4_2026-03-05", "month": "202605", "reasoning_effort": "medium", "min_score_filter": null, "total_candidates": 95, "total_tested": 95, "timestamp": "2026-07-23T23:11:12", "input_dir": "/mnt/shared/hanze/local/icml2026-research-level-math/Mathematician/helpers/qa...
{ "ge5_hard": { "correct": 46, "total": 95, "accuracy": 0.4842105263157895 } }
{ "correct": 46, "total": 95, "accuracy": 0.4842105263157895 }
[ { "id": "2605.24392v1::orig", "theorem_type": "ge5_hard", "score": 5, "correct_answer": "A", "model_answer": "A", "is_correct": true, "reasoning_effort": "medium", "requested_reasoning_effort": "medium", "instance": "msra/shared", "elapsed_seconds": 14.587110757827759, "e...
{ "date": "0724_042756", "model": "gpt-5.4_2026-03-05", "month": "202606", "reasoning_effort": "medium", "min_score_filter": null, "total_candidates": 116, "total_tested": 116, "timestamp": "2026-07-24T04:59:05", "input_dir": "/mnt/shared/hanze/local/icml2026-research-level-math/Mathematician/helpers/...
{ "ge5_hard": { "correct": 49, "total": 116, "accuracy": 0.4224137931034483 } }
{ "correct": 49, "total": 116, "accuracy": 0.4224137931034483 }
[ { "id": "2606.27329v2::alt1", "theorem_type": "ge5_hard", "score": 6, "correct_answer": "B", "model_answer": "B", "is_correct": true, "reasoning_effort": "medium", "requested_reasoning_effort": "medium", "instance": "gcr/shared", "elapsed_seconds": 12.666792392730713, "er...

LiveMathematicianBench v7 — arXiv 2026-03 … 2026-06

An automated, refreshable benchmark of research-level mathematics multiple-choice questions, generated from newly published arXiv math papers (contamination-resistant by construction). Columbia University & Microsoft Research.

Each question is grounded in a theorem from a recent paper; distractors are adversarially crafted from the proof sketch, and a multi-stage hardness pipeline keeps the final set difficult for frontier models.

Layout

Per month (202603/, 202604/, 202605/, 202606/):

  • full/qaEval_<month>_full.json — all generated MCQs (v5)
  • ge5/qaEval_<month>_ge5.json — quality-filtered (rubric ALS+TAS+GPS+DQS total_score ≥ 5)
  • hard/qaEval_<month>_ge5_hard.json — final ge5-hard subset (one hardest item per source theorem)
  • hard/accuracy_test_<month>_medium_filter2.json — final independent retest results
  • summary.json — per-month funnel

Results — GPT-5.4 (medium) final hard-set accuracy

Month hard questions accuracy
2026-03 112 37.5%
2026-04 108 42.6%
2026-05 95 48.4%
2026-06 116 42.2%
Overall 431 42.5%

Lower accuracy = harder. See summary.json for the full generation → ge5 → stem-nontrivial → hard funnel.

Downloads last month
77