HerrHruby/acemath_rl_hard
收藏Hugging Face2025-10-13 更新2025-10-25 收录
下载链接：
https://hf-mirror.com/datasets/HerrHruby/acemath_rl_hard
下载链接
链接失效反馈官方服务：
资源简介：
该数据集基于CohenQu/AceReason-Math-Qwen3-4B数据集进行处理，处理步骤包括：过滤掉平均奖励大于或等于0.7的样本，以及对平均奖励为0.0的样本进行降采样。README文件中未提供数据集的直接描述。
The dataset is based on the CohenQu/AceReason-Math-Qwen3-4B dataset and processed by filtering out samples with a mean reward of 0.7 or higher, and downsampling samples with a mean reward of exactly 0.0. No direct description of the dataset is provided in the README.
提供机构：
HerrHruby
