Dataset · EleutherAI
MATH math problem dataset
EleutherAI/hendrycks_math
10k+ math problems with solutions for training and evaluating reasoning models.
What's inside
A collection of text-based math problems from the MATH benchmark (Hendrycks et al., NeurIPS 2021). The card specifies a size of 10k to 100k examples; data is sourced from the open Hendrycks/math repository. Format and language are not specified in the card.
What you can do with it
- Train a model to solve math problems
- Evaluate LLM reasoning capabilities on the MATH benchmark
- Create an SFT dataset for step-by-step solutions
- Size
- 10K–100K rows
- Downloads this month
- 169.7K
- Likes
- 121
Models trained on it
- Ternii/Qwen3.6-35B-A3B-TR2.25-G-GGUFText generation 1.4K
- zaydsyed/TactLM-Math-ProcessGRPO-Qwen2.5-0.5BText generation 1K
- mradermacher/Atem-SageMaths-1.5B-i1-GGUF 617
- Ternii/Qwen3.8-Flash-Next-TR2.31-G-GGUFText generation 617
- Rinstitute/Rina-Juvenile-Exp-2B-GGUFText generation 547
- Ternii/Qwen3.8-Flash-Next-TR3-G-GGUFText generation 475
- mradermacher/sarashina-13B-finetuned-v1-i1-GGUF 468
- Ternii/Qwen3.6-27B-TR2.5-GText generation 468
How to load it
from datasets import load_dataset
ds = load_dataset("EleutherAI/hendrycks_math")