What’s Out There
← Back to swiping
EducationFreeNonprofit

NuminaMath-CoT

Project Numina

About 860,000 math problems with chain-of-thought solutions, from Chinese high school to olympiad level.

Problems collected from exam PDFs and math discussion forums, with solutions rewritten into chain-of-thought format by a model. One row is a problem with its solution text. It trained the winning model of the first AI Mathematical Olympiad progress prize.

Access

Cost
FreeNo charge.
License
Open licenseApache 2.0
Sign-up
No sign-up
How you get it
API and Bulk downloadParquet

What's in it

One row is
One row per problem and solution
Covers
Not geographic
Years
2024
Updated
One-time release
Size
About 860,000 problems
problemsolutionsourcemessages

Quality

3 of 5

Broad coverage, but solutions were rewritten by a model and some may be wrong.

  • Source problems from exams and forums have unclear rights
  • Solution quality varies by source
  • Possible overlap with common math benchmarks

Connects to

  • OpenR1-Math-220k: Takes problems from NuminaMath 1.5 and adds R1 traces
  • MATH-500: Benchmark whose style these problems resemble; check for overlap before evaluating

Questions it could answer

  • Which sources have the highest share of unsolved or wrong solutions?
  • How does problem difficulty distribute by source?

Similar data sets

  • OpenMathReasoning

    306,000 forum math problems with 3.2 million long chain-of-thought solutions.

    EducationNVIDIAFree

  • OpenR1-Math-220k

    220,000 math problems, each with two to four DeepSeek-R1 reasoning traces, many checked automatically.

    EducationHugging Face (Open-R1 project)Free

  • Llama-Nemotron Post-Training Dataset

    Millions of math, code, and reasoning examples used to post-train NVIDIA's Llama-Nemotron models.

    Language & textNVIDIAFree

  • OpenThoughts-114k

    114,000 long reasoning traces across math, science, code, and puzzles, released openly.

    Language & textOpen Thoughts teamFree

  • FineMath

    34 billion tokens of math-teaching web pages, chosen by a classifier trained on Llama annotations.

    EducationHugging Face (SmolLM team)Free

  • MMLU-Pro

    A harder MMLU with ten answer options and 12,000 reasoning-heavy questions.

    EducationTIGER-Lab (University of Waterloo and collaborators)Free