budecosystem/LMArena-Human-Preference-55K-Complexity-train

该数据集包含了模型A和模型B对特定提示（prompt）的响应数据，以及判断哪个模型响应更优的信息。每个样本包含唯一的ID，两个模型的名称，提示文本，两个模型的响应文本，获胜模型的标记（模型A、模型B或平局），以及一个表示问题复杂度的浮点数。数据集分为训练集，共有57477个样本。

The dataset consists of responses from Model A and Model B to specific prompts, along with information on which models response is considered better. Each sample includes a unique ID, the names of the two models, the prompt text, the response texts from both models, a label indicating the winning model (Model A, Model B, or a tie), and a floating-point number representing the complexity of the issue. The dataset is split into a training set, which contains a total of 57477 samples.