sparsh35/filtered_lm_arena

该数据集包含多个模型对比的结果，每个样本包括两个模型的答案（answer_a和answer_b），以及标识哪个模型答案胜出的字段（winner）。数据集还包含模型名称（model_a和model_b）、语言（language）、提示信息（prompt）和唯一标识符（id）。训练集包含大约13.56万个样本。

The dataset contains results comparing multiple models, with each sample including answers from two models (answer_a and answer_b), and a field indicating which models answer is the winner (winner). The dataset also includes model names (model_a and model_b), language (language), prompt information (prompt), and a unique identifier (id). The training set contains approximately 135,600 samples.