Issues

• MathJudger accepts tuples and intervals with extra or missing components, e.g. (2,4,6) for gold (2,4)

  OpenBMB/OlympiadBench#21 · Filter by author damienen opened on Nov 2, 2026

• inference/judge.py passes the model answer as the ground truth, so per-answer tolerances follow the model's answer order

  OpenBMB/OlympiadBench#20 · Filter by author damienen opened on Nov 2, 2026

• 牛逼

  OpenBMB/OlympiadBench#19 · Filter by author albert-jin opened on Apr 19, 2026

• 牛逼

  OpenBMB/OlympiadBench#18 · Filter by author albert-jin opened on Apr 19, 2026

• Are there results for more recent Language Models; or potential data contamination detection?

  OpenBMB/OlympiadBench#17 · Filter by author JingzheShi opened on Jul 28, 2025

• Final Answer needs update for question 1618

  OpenBMB/OlympiadBench#15 · Filter by author lulu5049 opened on May 22, 2025

• Code bug in text-only GPT-4 evaluation

  OpenBMB/OlympiadBench#11 · Filter by author sudanl opened on Jan 14, 2025

• 【Eval】Whats the difference between "eval/auto_scoring_judge" and "inference/code/judge.py"?

  OpenBMB/OlympiadBench#10 · Filter by author Violettttee opened on Oct 10, 2024

• 对评估的结果有些疑问

  OpenBMB/OlympiadBench#8 · Filter by author qizheyanger opened on Aug 6, 2024