selfcorrexp2/w2r125k_r2r115k_r100k

该数据集包含多个字段，如索引(idx)，提示(prompt)，是否为第一轮(first_round)，正确答案(gt)等。部分字段如rewards和my_solu是序列类型，表明可能包含列表或文本序列。数据集还包含对话信息，其中有内容(content)和角色(role)。此外，数据集还包含难度等级(level)，类型(type)，解决方案(solution)和预测(pred)等字段。数据集分为训练集(train)，共有283,126个示例，大小约为1.7GB。

The dataset includes fields such as index (idx), prompt, whether it is the first round (first_round), ground truth (gt), etc. Some fields like rewards and my_solu are of sequence type, indicating they may contain list or text sequences. The dataset also includes conversation information with content and role. In addition, the dataset contains fields like difficulty level (level), type (type), solution, and prediction (pred). The dataset is split into a training set (train) with a total of 283,126 examples, totaling approximately 1.7GB in size.