AmirMohseni/arena-preference-data-no-ties

该数据集包含对话信息，每个对话有多个特征，包括对话双方的发言内容、角色、使用的模型、赢家信息、对话的元数据（如发言的token数、标题和列表的使用次数等），以及对话的分类标签（如创造性写作、数学问题等）。数据集分为全部数据、训练集和测试集三个部分。

The dataset contains conversation information, with each conversation having multiple features including the speech content and roles of both parties, the models used, winner information, metadata about the conversation (such as the number of tokens spoken, the usage count of headers and lists, etc.), and categorical tags for the conversation (such as creative writing, math problems, etc.). The dataset is divided into three parts: all data, training set, and test set.