selfcorrexp2/w2r125k_r2r115k_r125k

这是一个包含多个字段的数据集，字段类型包括整数、字符串、布尔值以及序列和列表。具体包括索引、提示、是否为第一轮、真实标签、奖励、解决方案、预测结果等字段。另外，还包括对话内容列表，对话列表中又包括内容和角色两个字段。数据集被划分为训练集，并提供了相应的配置信息。

This dataset contains multiple fields with various data types including integers, strings, booleans, sequences, and lists. It includes fields such as index, prompt, whether its the first round, ground truth, rewards, solution, prediction, etc. Additionally, it has a list of conversation contents, within which there are two sub-fields: content and role. The dataset is split into a training set and provides corresponding configuration information.