metadata
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.