selfcorrexp2/fixed_llama3_morecorr_sft_4dpo_type1_5ktype2_7k_type3_8k_type4_ver2

该数据集包含文本对和标签信息，用于某种文本选择或评估任务。具体来说，它包括选中的文本(chosen_txt)、被拒绝的文本(rejected_txt)、真实标签(gt)、选中的标识(chosen)、被拒绝的标识(rejected)、提示文本(prompt)以及一个名为margin的浮点数，可能是用于评估选择的度量。数据集分为训练集(train)，包含约35万个示例。数据集的总大小约为341MB。The dataset contains pairs of texts and label information, used for some text selection or evaluation task. Specifically, it includes the chosen text (chosen_txt), the rejected text (rejected_txt), the ground truth label (gt), the chosen indicator (chosen), the rejected indicator (rejected), the prompt text (prompt), and a floating-point number named margin, which may be used for evaluating the selection. The dataset is split into a training set (train) with approximately 350,000 examples. The total size of the dataset is about 341MB.