---
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.