jeevana28/nectar_llm_judge

这是一个包含对话数据的数据集，其中包括提示信息(prompt)、模型输出(model1和model2)、响应(response1和response2)、排名(rank1和rank2)、对话轮数(turns)、是否友善(good_natured)、数据来源(source)、索引(index)和分数(score)等字段。数据集分为训练集和测试集，共包含84000个训练样本和21000个测试样本。

This is a dataset containing dialogue data, which includes fields such as prompt, model outputs (model1 and model2), responses (response1 and response2), rankings (rank1 and rank2), number of turns in the conversation (turns), friendliness (good_natured), data source (source), index, and score. The dataset is divided into training and test sets, containing a total of 84,000 training samples and 21,000 test samples.