OpenMathReasoning nvidia

OpenMathReasoning OpenMathReasoning is a large-scale math reasoning dataset for training large language models (LLMs). This dataset contains 306K unique mathematical problems sourced from AoPS forums with: 3.2M long chain-of-thought (CoT) solutions 1.7M long tool-integrated reasoning (TIR) solutions 566K samples that select the most promising solution out of many candidates (GenSelect) Additional 193K problems sourced from AoPS forums (problems only, no solutions) We used… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/OpenMathReasoning.

Type
dataset
License
cc-by-4.0
Language
en
Downloads
45,655
Likes
470
Access
public
Files
0

查看完整页面 · 查看原文