NotDevinLe/prism

该数据集包含了对话信息，每个对话示例包括提示内容（包括内容和角色）、用户选择的内容、被拒绝的内容、用户标识、对话ID、当前轮次编号和总轮次数。数据集分为训练集（分为已见和未见部分）和测试集（分为已见和未见部分），用于训练和评估对话系统。

The dataset contains conversational information, with each example including prompt content (including text and role), user-chosen content, rejected content, user ID, dialog ID, current turn number, and total number of turns. The dataset is split into training sets (seen and unseen) and test sets (seen and unseen) for training and evaluating dialogue systems.