saepark/hh-rlhf-single-turn-RM-train-furthersplit-RM-train-actual-after-sleeper-preference

这是一个包含对话数据的训练集，每个示例包括一个提示(prompt)、提示ID(prompt_id)、选定的回复(chosen)、被拒绝的回复(rejected)、对话消息(messages)、选定回复的得分(score_chosen)和被拒绝回复的得分(score_rejected)。选定的回复和被拒绝的回复都包含回复内容和角色信息。此外，每个示例还包含其他信息，如数据来源(source)。

This is a training dataset containing conversational data, with each example including a prompt, prompt ID, chosen response, rejected response, conversation messages, score of the chosen response, and score of the rejected response. Both the chosen and rejected responses contain response content and role information. Additionally, each example also includes other information such as the data source (source).