datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
MATH-500
Dataset Card for MATH-500
This dataset contains a subset of 500 problems from the MATH benchmark that OpenAI created in their Let's Verify Step by Step paper. See their GitHub repo for the source file: https://github.com/openai/prm800k/tree/main?tab=readme-ov-file#math-splits
aime25
AIME 25
American Invitational Mathematics Examination (AIME) 2025
Citation
If you use the AIME25 dataset in your research, please consider citing it as follows:
@misc{aime25,
title={American Invitational Mathematics Examination (AIME) 2025},
author={Zhang, Yifan and Math-AI, Team},
year={2025},
}
MetaMathQAView the project page:
https://meta-math.github.io/
see our paper at https://arxiv.org/abs/2309.12284
Note
All MetaMathQA data are augmented from the training sets of GSM8K and MATH.
None of the augmented data is from the testing set.
You can check the original_question in meta-math/MetaMathQA, each item is from the GSM8K or MATH train set.
Model Details
MetaMath-Mistral-7B is fully fine-tuned on the MetaMathQA datasets and based on the powerful Mistral-7B model. It is… See the full description on the dataset page: https://huggingface.co/datasets/meta-math/MetaMathQA.aime26
AIME 26
American Invitational Mathematics Examination (AIME) 2026
Citation
If you use the AIME26 dataset in your research, please consider citing it as follows:
@misc{aime26,
title={American Invitational Mathematics Examination (AIME) 2026},
author={Zhang, Yifan and Math-AI, Team},
year={2026},
}
Omni-MATH
Dataset Card for Omni-MATH
Recent advancements in AI, particularly in large language models (LLMs), have led to significant breakthroughs in mathematical reasoning capabilities. However, existing benchmarks like GSM8K or MATH are now being solved with high accuracy (e.g., OpenAI o1 achieves 94.8% on MATH dataset), indicating their inadequacy for truly challenging these models. To mitigate this limitation, we propose a comprehensive and challenging benchmark specifically designed… See the full description on the dataset page: https://huggingface.co/datasets/KbsdJames/Omni-MATH.dapo-math-17kminervamathMathInstruct
🦣 MAmmoTH: Building Math Generalist Models through Hybrid Instruction Tuning
MathInstruct is a meticulously curated instruction tuning dataset that is lightweight yet generalizable. MathInstruct is compiled from 13 math rationale datasets, six of which are newly curated by this work. It uniquely focuses on the hybrid use of chain-of-thought (CoT) and program-of-thought (PoT) rationales, and ensures extensive coverage of diverse mathematical fields.
Project Page:… See the full description on the dataset page: https://huggingface.co/datasets/TIGER-Lab/MathInstruct.swallow-math-v2
SwallowMath-v2
Resources
📑 arXiv: Read our paper for detailed methodology at arXiv:2505.02881.
🤗 Sister Dataset: Discover SwallowCode2, our companion dataset for code generation.
🧮 What is it?
SwallowMath-v2 is a large-scale mathematical dataset containing 32 billion tokens, developed as the successor to SwallowMath-v1.
Building on the success of v1, this release aims to construct a larger-scale and more permissively licensed corpus to support open and… See the full description on the dataset page: https://huggingface.co/datasets/tokyotech-llm/swallow-math-v2.math-contests-2026
Math Contests 2026 (🔗 notadib/math-contests-2026)
197 problems from national olympiads and team-selection tests held January 2026 and onward — a held-out benchmark for math reasoning, sourced after the contests ran but before solutions were widely propagated, so they should not appear in any current LLM training data.
Excluded: any contest held in 2025 — BMO Round 1 (Nov 2025), USA TSTST, USA TST (Dec 2025) and Bundeswettbewerb Mathematik (Dec 2025) — kept strictly to events… See the full description on the dataset page: https://huggingface.co/datasets/notadib/math-contests-2026.MetaMathQA-40Karxiv.org/abs/2309.12284
View the project page:
https://meta-math.github.io/
Nemotron-Math-Proofs-v3-SFT
Nemotron-Math-Proofs-v3-SFT
Dataset Description:
Nemotron-Math-Proofs-v3-SFT is a long-form mathematical reasoning dataset containing proof-generation, proof-refinement, verification, and meta-verification traces. The release contains 414,890 samples representing 15,818 unique problems after quality filtering.
The source pool contains 15,879 hard proof problems selected from the AoPS subset of nvidia/Nemotron-Math-Proofs-v1. Responses are generated using… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/Nemotron-Math-Proofs-v3-SFT.formal-math-autoformalization
Formal Math Autoformalization Dataset
A growing, CC0 public-domain corpus of ⟨natural-language statement ↔ Lean 4 statement + proof⟩ pairs, contributed through the Agentic Commons network.
Why this is scarce data. Mathlib already contains millions of proven Lean theorems — but as bare Lean, with no paired natural language:
theorem add_comm (a b : ℕ) : a + b = b + a := ... -- no "addition on naturals is commutative" attached
The scarce, valuable artifact is the pairing of the… See the full description on the dataset page: https://huggingface.co/datasets/AgenticCommons/formal-math-autoformalization.open-math-courses
Open Mathematics Courses
Graduate and research-level mathematics lessons with complete proofs, worked examples and solved exercises, one row
per lesson. The lessons are the Markdown sources of the public site https://kokunoyumeto.github.io/open-math-courses-public/ (snapshot commit 4219d3885c68).
3,145 lessons in 140 courses, about 16,927 thousand words.
Fields: course_id, course_title, lesson_title, source_path, page_url, authorship, license, words,
text (Markdown with TeX… See the full description on the dataset page: https://huggingface.co/datasets/KokunoYumeto/open-math-courses.math-tiers
Math-Tiers: A Tiered Pretraining Corpus for Studying Numerical Reasoning
A large-scale English pretraining corpus split into three tiers by mathematical content density. Designed for controlled experiments studying how data composition during pretraining affects numerical reasoning in language models.
Tiers
Tier
Description
Shards
Size
Est. Tokens
Sources
T0
Pure narrative: no digits, number words, or math
648
542 GB
~113B
RedPajama-Book, PleIAs/English-PD… See the full description on the dataset page: https://huggingface.co/datasets/swan07/math-tiers.hle-mathGSM8KInstruct_ParallelMATH-Hard
Dataset Card for Mathematics Aptitude Test of Heuristics, hard subset (MATH-Hard) dataset
Dataset Summary
The Mathematics Aptitude Test of Heuristics (MATH) dataset consists of problems
from mathematics competitions, including the AMC 10, AMC 12, AIME, and more.
Each problem in MATH has a full step-by-step solution, which can be used to teach
models to generate answer derivations and explanations. For MATH-Hard, only the
hardest questions were kept (Level 5).… See the full description on the dataset page: https://huggingface.co/datasets/lighteval/MATH-Hard.StackMathQA
StackMathQA
StackMathQA: A Curated Collection of 2 Million Mathematical Questions and Answers Sourced from Stack Exchange
StackMathQA is a meticulously curated collection of 2 million mathematical questions and answers, sourced from various Stack Exchange sites. This repository is designed to serve as a comprehensive resource for researchers, educators, and enthusiasts in the field of mathematics and AI research.
Configs
configs:
- config_name: stackmathqa1600k… See the full description on the dataset page: https://huggingface.co/datasets/math-ai/StackMathQA.MathV360K
Overview
MathV360K is proposed by Math-LLaVA: Bootstrapping Mathematical Reasoning for Multimodal Large Language Models, which consists 40K images from 24 datasets and 360K question-answer pairs. MathV360K is used to enhance the multimodal mathematical reasoning capabilities of MLLMs, achieving 46.6% accuracy on MathVista benchmark and 15.69% accuracy on MathVision dataset.
Paper or resources for more information:
[Paper] [Code] [Model]
Source Data
TAL-SCQ5KTAL-SCQ5K
Dataset Description
Dataset Summary
TAL-SCQ5K-EN/TAL-SCQ5K-CN are high quality mathematical competition datasets in English and Chinese language created by TAL Education Group, each consisting of 5K questions(3K training and 2K testing). The questions are in the form of multiple-choice and cover mathematical topics at the primary,junior high and high school levels. In addition, detailed solution steps are provided to facilitate CoT training and all the… See the full description on the dataset page: https://huggingface.co/datasets/math-eval/TAL-SCQ5K.MathReal
Dataset Card for MathReal
Dataset Description
Paper Information
Dataset Examples
Leaderboard
Citation
Dataset Description
The MathReal dataset is designed to evaluate the performance of Multi-modal Large Language Models (MLLMs)on real-world K-12 mathematical questions. It consists of 2,000 high-quality math problems, each represented as an image captured in authentic educational contexts. The dataset includes various types of questions, such as multiple-choice… See the full description on the dataset page: https://huggingface.co/datasets/junfeng0288/MathReal.mathmetics-dataset-intmax
Transformer Math Dataset (200,000,000 Samples Sharded)
High-precision synthetic mathematical expression dataset generated for training sequence-to-sequence math Transformers in JAX/Flax.
Dataset Structure
Total Samples: 200,000,000
Shard Format: JSONL sharded files (100,000 samples per shard)
Supported Operations: +, -, *, /, ^, sin, cos, tan, log, ln, exp, sqrt, abs
Expression Depth Range: Depth 4 to 6
Integer Operand Ratio: 80%
Data Fields
Each… See the full description on the dataset page: https://huggingface.co/datasets/saidurga001301/mathmetics-dataset-intmax.math-pretraining-corpusAceReason-Math
AceReason-Math Dataset
Overview
AceReason-Math is a high quality, verfiable, challenging and diverse math dataset for training math reasoning model using reinforcement leraning. This dataset contains
49K math problems and answer sourced from NuminaMath and DeepScaler-Preview
applying filtering rules to exclude unsuitable data (e.g., multiple sub-questions, multiple-choice, true/false, long and complex answers, proof, figure)
this dataset was used to train… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/AceReason-Math.GSM8K_zh
Dataset
GSM8K_zh is a dataset for mathematical reasoning in Chinese, question-answer pairs are translated from GSM8K (https://github.com/openai/grade-school-math/tree/master) by GPT-3.5-Turbo with few-shot prompting.
The dataset consists of 7473 training samples and 1319 testing samples. The former is for supervised fine-tuning, while the latter is for evaluation.
for training samples, question_zh and answer_zh are question and answer keys, respectively;
for testing samples, only… See the full description on the dataset page: https://huggingface.co/datasets/meta-math/GSM8K_zh.swallow-math
SwallowMath
October 21, 2025: Newer versions are available: SwallowCode-v2 and SwallowMath-v2 have been released with improved rewriting pipelines.
Resources
🐙 GitHub: Explore the project repository, including pipeline code and prompts at rioyokotalab/swallow-code-math.
📑 arXiv: Read our paper for detailed methodology and results at arXiv:2505.02881.
🤗 Sister Dataset: Discover SwallowCode, our companion dataset for code generation.
What is it?… See the full description on the dataset page: https://huggingface.co/datasets/tokyotech-llm/swallow-math.MGSM8KInstruct_CrossMathOdysseyNemotron-Math-Proofs-v1
Nemotron-Math-Proofs-v1
Paper: Nemotron-Math: Efficient Long-Context Distillation of Mathematical Reasoning from Multi-Mode SupervisionCode: https://github.com/NVIDIA/NeMo-SkillsDocumentation: Nemotron-MathProofs-v1 documentation
Dataset Description:
Nemotron-Math-Proofs-v1 is a large-scale mathematical reasoning dataset containing ~580k natural language proof problems, ~550k formalizations into theorem statements in Lean 4, and ~900k model-generated reasoning… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/Nemotron-Math-Proofs-v1.