selfcorrexp2/w2r125k_r2r60k_r150k

该数据集包含了对话的索引、提示、是否为第一轮对话、真实标签、奖励信息、解决方案、预测结果等字段。数据集中的对话被分为不同的角色和内容，适用于对话生成或评估任务。数据集分为训练集，其大小为约2GB，共有约333,126个示例。

The dataset includes fields such as conversation index, prompt, whether it is the first round of the conversation, ground truth label, reward information, solution, prediction results, etc. The conversations in the dataset are categorized by role and content, suitable for dialogue generation or evaluation tasks. The dataset is split into a training set, which is about 2GB in size and contains approximately 333,126 examples.