open-r1/DAPO-Math-17k-Processed
收藏官方服务：
资源简介：
这是一个经过处理的数据集，是DAPO-Math-17k的版本，我们去重了提示，并将提示和地面真实答案格式化以兼容TRL的GRPO训练器。我们还派生出了纯英文和纯中文的子集。
This is a processed version of the DAPO-Math-17k dataset where we have deduplicated the prompts and reformatted the prompts and ground truth answers to be compatible with TRLs GRPO trainer. We have also derived pure English and Chinese subsets.
提供机构：
open-r1
