gupta-tanish/Ultrafeedback-DPOx4C2

该数据集包含了用于训练和测试的文本数据，其中包括提示信息(prompt)、提示ID(prompt_id)、选中的文本及其角色(chosen)、被拒绝的文本及其角色(rejected)、对话消息(messages)以及选中文本和拒绝文本的评分(score_chosen和score_rejected)。数据集分为训练和测试两个部分，每个部分又细分为偏好(train_prefs/test_prefs)、敏感度(train_sft/test_sft)和生成(train_gen/test_gen)三个子集。

The dataset consists of text data for training and testing, including prompt information (prompt), prompt ID (prompt_id), selected text and its role (chosen), rejected text and its role (rejected), conversation messages (messages), and scores for the chosen and rejected texts (score_chosen and score_rejected). The dataset is divided into training and testing parts, each of which is further subdivided into preference (train_prefs/test_prefs), sensitivity (train_sft/test_sft), and generation (train_gen/test_gen) subsets.