Skip to content

Dataset · HuggingFaceH4

MATH-500 benchmark task subset

HuggingFaceH4/MATH-500

A subset of 500 math problems from the MATH benchmark for evaluating step-by-step reasoning.

What's inside

The dataset contains 500 problems from the MATH benchmark created by OpenAI in the work Let's Verify Step by Step. This is a small text subset in English; the original file is available in the OpenAI prm800k repository. The data is intended for evaluating models' mathematical reasoning.

What you can do with it

  • Evaluate LLMs on mathematical problems
  • Assess chain-of-thought and step-by-step reasoning
  • Use as a compact reasoning benchmark
Tasks
Text generation
Languages
English
Size
up to 1K rows
Downloads this month
218.4K
Likes
358

How to load it

from datasets import load_dataset

ds = load_dataset("HuggingFaceH4/MATH-500")