Bmingg/DPO_3a_llama_3600_constrained_mbr_bleu_vs_ancestral_sampling_bleurt_strict_bleu

该数据集包含三个字段：prompt（提示）、chosen（选择的回答）和rejected（被拒绝的回答）。它似乎用于某种文本选择任务，其中prompt可能是问题或情境，而chosen和rejected是可能的回答选项。训练集包含3600个示例，数据集总大小为1,784,158字节。The dataset includes three fields: prompt (a cue or situation), chosen (the selected response), and rejected (the response that was not chosen). It appears to be used for some kind of text selection task, where the prompt might be a question or situation, and chosen and rejected are possible response options. The training set contains 3,600 examples, and the total size of the dataset is 1,784,158 bytes.