Skip to main content

> ML_DATASET // HENDRYCKS-MATH-DATASET_v1.0

MATH: Measuring Mathematical Problem Solving (Hendrycks et al. 2021)

UC Berkeley (Dan Hendrycks et al.) · Advanced Mathematical Reasoning · 12,500 high school competition math problems (AMC 10, AMC 12, AIME)

Advanced Mathematical ReasoningMIT12,500 high school competition math problems (AMC 10, AMC 12, AIME)open

Dataset Profile & Characteristics

Label Type:Step-by-step LaTeX solution rationale and boxed final answer across 7 subjects and 5 difficulty levels
Languages:en
License Tier:permissive-open-source
Modalities:text

Intended Use

  • Evaluating deep chain-of-thought and reinforcement-learning mathematical reasoning in frontier LLMs

Prohibited / Discouraged Use

  • General reading comprehension assessment

Bias, Leakage & Privacy Risk Analysis

Privacy / Sensitive Data Risks:

Math competition problems; zero personal data.

Known Bias:

US high school mathematics competition curriculum formatting.

Known Benchmark Leakage:

Heavy contamination risk in modern web-scale pretraining datasets.

Compatible Tools & Libraries