selfcorrexp2/w2r100k_r2r40k_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)、真实标签(gt)、奖励(rewards)、用户解决方案(my_solu)、标记(flag)、轮次(turn)、对话(conversations，包括对话内容和角色)、难度等级(level)、类型(type)、解决方案(solution)、预测(pred)以及用户提示的对话(my_prompt_conversations)。数据集被划分为训练集，包含240000个示例，文件大小约为1.44GB。具体数据集的内容和用途在README中未描述。

The dataset includes multiple fields such as index (idx), prompt, whether its the first round (first_round), ground truth (gt), rewards, users solution (my_solu), flag, turn, conversation (including conversation content and role), difficulty level (level), type, solution, prediction (pred), and user prompt conversations (my_prompt_conversations). The dataset is split into a training set with 240,000 examples, and the file size is approximately 1.44GB. The specific content and purpose of the dataset are not described in the README.