selfcorrexp2/Self-rewarding-non-Delete-Llama3-tmp10-generation

这个数据集包含了索引（idx）、提示文本（prompt）、答案（answers）、对话内容（conversations）、代理奖励（proxy_rewards）、真实标签（gt）和真实奖励（true_rewards）等字段。数据集分为训练集（train），共有15000个示例，大小为53157738字节。配置部分提供了默认配置和数据文件的路径。

The dataset includes fields such as index (idx), prompt text (prompt), answers (answers), conversation content (conversations), proxy rewards (proxy_rewards), ground truth labels (gt), and true rewards (true_rewards). The dataset is split into a training set (train) with a total of 15,000 examples and a size of 53,157,738 bytes. The configuration section provides the default configuration and the path to the data files.