ZixuanKe/bopang_sup_dpo_binarized_train_chunk_1

该数据集包含用于训练的数据，主要特征包括提示（prompt）、选择的真实答案（ground_truth_chosen）、以及两个被拒绝的版本（v1_rejected和v1_1_rejected）。数据集包含一个训练分割，共有1,077,379个例子，数据文件路径为data/train-*。This dataset contains training data with main features including prompt, ground_truth_chosen, and two rejected versions (v1_rejected and v1_1_rejected). The dataset includes a training split with a total of 1,077,379 examples, and the data file path is data/train-*.