Skip to content

Dataset · SevenAIGmbH

MATH in lighteval format for LLM evaluation

SevenAIGmbH/MATH-lighteval

Math competition problems with step-by-step solutions for benchmarking in lighteval.

What's inside

A dataset of mathematical problems from AMC 10, AMC 12, AIME, and other competitions. Each problem has a complete step-by-step solution, allowing models to learn how to derive answers and explanations. This is a MATH version in lighteval format with ready-made builder configs; language is English, volume is 10 to 100k examples.

What you can do with it

  • Evaluating mathematical reasoning in lighteval
  • Fine-tuning on step-by-step solutions
  • Comparing heuristics and models on MATH
Languages
English
Size
10K–100K rows
Downloads this month
35.8K
Likes
67

How to load it

from datasets import load_dataset

ds = load_dataset("SevenAIGmbH/MATH-lighteval")