| pretty_name: Math Evaluation Benchmarks | |
| language: | |
| - en | |
| task_categories: | |
| - question-answering | |
| configs: | |
| - config_name: default | |
| data_files: | |
| - split: validation | |
| path: data/val.parquet | |
| # Math Evaluation Benchmarks | |
| A collection of 273 competition-math problems for evaluating mathematical | |
| reasoning. It contains AMC23, AIME24, AIME25, AIME26, BRUMO25, HMMT February | |
| 2025, and CMIMC25. | |
| Each example contains a chat-formatted prompt, the ground-truth final answer, | |
| the benchmark name in `data_source`, and basic example metadata. | |