AmirMohseni/arena-preference-data-filtered

该数据集包含对话信息，每个对话包括选中的(chosen)和拒绝的(rejected)内容，每种内容都有其内容和角色(role)。此外，还包括对话ID，模型A和模型B的标识，获胜模型，以及关于对话的一些元数据，如每个助手的token数量、标题和列表的数量等。数据集还包含分类标签，如创意写作、数学问题等，以及语言类型。数据集分为验证集和测试集。

The dataset contains conversation information, including chosen and rejected content with their respective roles. It also includes conversation ID, identifiers for model A and model B, the winning model, and metadata about the conversation such as the number of tokens for each assistant, the count of headers and lists, etc. The dataset also has category tags like creative writing, math problems, etc., and the language type. The dataset is split into validation and test sets.