preference-team/dataset-for-annotation-v2-annotated

该数据集包含会话信息，每个会话由一个指令和两个可能的响应组成，以及参与者对这两个响应的偏好。数据集中的字段包括会话ID、消息内容、角色（可能是发出指令或响应的人）、指令文本、两个响应文本、参与模型1和模型2的名称、原始ID和用户偏好的分数。数据集分为训练集，大小为8248012字节，共有1143个示例。

The dataset contains conversational information, with each conversation consisting of an instruction and two possible responses, along with the participants preference for these responses. The fields in the dataset include session ID, message content, role (possibly the issuer of the instruction or the responder), instruction text, two response texts, the names of the participating models 1 and 2, the original ID, and the score of user preference. The dataset is split into a training set, which is 8248012 bytes in size and contains 1143 examples.