selfcorrexp2/llama3_sft_balanced_gen2_math_type4

该数据集包含文本选择任务的相关数据，其中包括选中的文本(chosen_txt)、被拒绝的文本(rejected_txt)、真实标签(gt)、选中的标记(chosen)、被拒绝的标记(rejected)、提示文本(prompt)以及一个表示差异的浮点数字段(margin)。数据集仅包含训练集，共有3153个示例。

The dataset contains data related to text selection tasks, including chosen text (chosen_txt), rejected text (rejected_txt), ground truth labels (gt), chosen markers (chosen), rejected markers (rejected), prompt text (prompt), and a floating-point field (margin) representing the difference. The dataset only contains a training set with a total of 3153 examples.