---
title: 'Measuring Mathematical Problem Solving With the MATH Dataset'
license: CC-BY-4.0
license_url: https://creativecommons.org/licenses/by/4.0/
license_note: 'paper is arXiv non-exclusive distribution license (verified on arXiv abs page 2026-11-12); this pointer note is our own words and is CC BY 4.0; we link to the paper rather than reproduce it'
source: https://iinventedlight.com/library/notes/2103-03874-math-dataset.md
source_license: 'arXiv non-exclusive distribution license'
attribution: 'Pointer by iinventedlight (AI-assisted). Source: Dan Hendrycks, Collin Burns, Saurav Kadavath, Akul Arora, Steven Basart, Eric Tang, Dawn Song, Jacob Steinhardt, Measuring Mathematical Problem Solving With the MATH Dataset (arXiv non-exclusive distribution license). This note links to the paper rather than reproducing it.'
author: 'iinventedlight (AI-assisted)'
date: 2026-11-12
status: published
approved_by: "Cole Withington, 2026-11-12 14:44 PT, Librarian chat widget t51s316: \"Yes, publish all 100 second-list arXiv notes\""
checked_on: 2026-11-12
as_of: 2026-11-12
confidence: medium
untested: 'Pointer note only. Paper not reproduced. Abs page and license checked 2026-11-12.'
category: arxiv
topic_area: 'ai-research: evals'
tags: arxiv, evals, mathematics, benchmarks
arxiv_id: 2103.03874
alias: 2103.03874-math-dataset
type: pointer
format: pointer
sources:
  - id: S1
    title: 'Measuring Mathematical Problem Solving With the MATH Dataset'
    publisher: arXiv
    authors: 'Dan Hendrycks, Collin Burns, Saurav Kadavath, Akul Arora, Steven Basart, Eric Tang, Dawn Song, Jacob Steinhardt'
    url: https://arxiv.org/abs/2103.03874
    version_read: 'v2 (8 Nov 2021)'
    accessed: 2026-11-12
    license: 'arXiv non-exclusive distribution license'
---
# Measuring Mathematical Problem Solving With the MATH Dataset

## Summary
[Measuring Mathematical Problem Solving With the MATH Dataset](https://arxiv.org/abs/2103.03874) (2021) released MATH, a set of 12,500 competition-style mathematics problems with full step-by-step solutions. It quickly became a standard test of mathematical reasoning for language models, sitting well above grade-school arithmetic sets in difficulty.

What it introduced: a large, carefully formatted competition-math benchmark whose solutions can also be used as training or few-shot rationales, so models can be scored on whether they can derive an answer.

What it led to: later hard math and science evals, and a long run of papers that treat MATH as a default reasoning target. Live notes: [2311-12022-gpqa](https://iinventedlight.com/library/notes/2311-12022-gpqa/) and [2501-14249-humanitys-last-exam](https://iinventedlight.com/library/notes/2501-14249-humanitys-last-exam/). This week's drafts also include [2411-04872-frontiermath](https://iinventedlight.com/library/notes/2411-04872-frontiermath/) and a pointer to [2410-05229-gsm-symbolic](https://iinventedlight.com/library/notes/2410-05229-gsm-symbolic/).

Use it when you want the original MATH dataset paper. Where to read it: https://arxiv.org/abs/2103.03874 (v2 (8 Nov 2021)).

The paper is posted on arXiv under the arXiv non-exclusive distribution license. This note links to it rather than reproducing it. The note itself is CC BY 4.0.

License: this note is CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Pointer by iinventedlight (AI-assisted). Paper license: arXiv non-exclusive distribution; this note links to it and does not reproduce it.

---

Agents: [/agents.md](/agents.md) | [/.well-known/agent-library.json](/.well-known/agent-library.json).
