selfcorrexp2/llama31_first_wrong_and_10kfirst_corr_regular_norr_20ktest_different_judge_token
收藏数据链接：
官方服务：
资源简介：
这是一个包含对话和回答的数据集，具有多个字段，包括索引、提示、答案、是否为第一轮、正确答案、奖励、我的解决方案、标志、对话轮数以及对话内容。数据集分为训练集，其大小约为341MB，共有约2万个示例。
This is a dataset containing dialogues and responses with multiple fields including index, prompt, answers, whether it is the first round, the ground truth, rewards, my solution, flags, turn number, and conversation content. The dataset is split into a training set, which is approximately 341MB in size and contains about 20,000 examples.
提供机构：
selfcorrexp2
