Synthetic Math Problem-Solution Pairs
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/RUCAIBox/JiuZhang3.0
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含了46亿个由GPT-4生成的数学问题-解答对，旨在提升较小规模的大型语言模型（LLM）的数学推理能力。通过提炼GPT-4的数据合成能力，生成了这个数据集，用于对九章3.0模型进行预训练，提供了高质量的训练数据，从而在各类数学推理任务中提升了模型的性能。该数据集的规模达到了46亿对，任务专注于数学推理训练。
This dataset comprises 4.6 billion math problem-solution pairs generated by GPT-4, designed to improve the mathematical reasoning abilities of smaller-scale large language models (LLMs). It was constructed by leveraging the data synthesis capabilities of GPT-4, and is employed for pre-training the Jiuzhang 3.0 model, providing high-quality training data to enhance the model's performance across a wide range of mathematical reasoning tasks. With a total of 4.6 billion pairs, this dataset focuses exclusively on mathematical reasoning training.
提供机构：
RUCAIBox
