gupta-tanish/Ultrafeedback-DPOx3

该数据集是一个包含对话和偏好的多字段数据集，用于训练和测试模型。数据集中的每个样本包含一个提示(prompt)，以及与之相关的选择的回复(chosen)和被拒绝的回复(rejected)，每个回复都包含内容和角色信息。此外，还有与对话相关的消息(messages)以及选择的回复和被拒绝的回复的分数。

The dataset is a multifield dataset containing dialogues and preferences, designed for training and testing models. Each sample in the dataset includes a prompt, along with the chosen and rejected responses associated with it, each containing content and role information. Additionally, there are messages related to the dialogue and scores for the chosen and rejected responses.