RLAIF/Value-v1-NUMINA-V1-Blocks-Merged-1620-problems-step-len-filtered
收藏Hugging Face2024-11-17 更新2025-04-08 收录
下载链接：
https://hf-mirror.com/datasets/RLAIF/Value-v1-NUMINA-V1-Blocks-Merged-1620-problems-step-len-filtered
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含了问题、解决方案、解决方案的步骤、即时回报（rtgs）、未折现的即时回报（undiscounted_rtgs）、是否正确、目标答案和解决方案计数的字段。数据集分为训练集和测试集，提供了默认配置的数据文件路径。
The dataset includes fields for problem, solution, solution steps, immediate rewards (rtgs), undiscounted immediate rewards (undiscounted_rtgs), correctness, target answer, and solution counts. It is split into training and test sets, and provides data file paths for the default configuration.
提供机构：
RLAIF
