selfcorrexp2/llama3_sft_balanced_corr_rr0k

该数据集包含了一系列特征，如索引、提示文本、是否为首次回合、真实标签、奖励序列、个人解决方案、标志位、回合数、会话内容和角色、等级、类型、预测序列和个人提示会话内容等。数据集被划分为训练集，其大小为1,562,883,602字节，包含249,308个样本。数据集配置中包含默认配置，指定了训练集的数据文件路径。

The dataset includes features such as index, prompt text, first round marker, ground truth label, reward sequence, personal solution, flag, turn number, conversation content and role, level, type, prediction sequence, and personal prompt conversation content. The dataset is split into a training set, which is 1,562,883,602 bytes in size and contains 249,308 samples. The dataset configuration includes a default setting that specifies the data file path for the training set.