selfcorrexp2/llama3_morewrong_rr60k_3ep_type3_8k_type4_5k

该数据集包含了选中的文本（chosen_txt）、被拒绝的文本（rejected_txt）、真实的标签（gt）、选择的标签（chosen）、拒绝的标签（rejected）、提示文本（prompt）和一个边缘值（margin）。数据集分为训练集，共有12574个样本。数据集的总大小约为84250007.16991222字节，下载大小为32020374字节。The dataset includes selected text (chosen_txt), rejected text (rejected_txt), ground truth labels (gt), chosen labels (chosen), rejected labels (rejected), prompt text (prompt), and a margin value. The dataset is split into a training set with a total of 12,574 samples. The total size of the dataset is approximately 84250007.16991222 bytes, with a download size of 32020374 bytes.