stochastic-parrots/rewardbench-lighteval

这是一个包含多种类型数据的综合数据集，适用于不同的自然语言处理任务。数据集的特征涵盖了文本提示、用户的选择和拒绝选项、数据子集、类别、子类别、选择的模型、拒绝的模型、唯一标识符和数据来源。数据集分为多个子集，包括测试、简单聊天、困难聊天、安全、推理等，每个子集都有特定数量的示例和大小。同时，提供了默认配置，指定了每个数据集分割的文件路径。

This is a comprehensive dataset containing various types of data suitable for different natural language processing tasks. The dataset features include text prompts, user choices and rejection options, subsets, categories, subcategories, chosen models, rejected models, unique identifiers, and data sources. The dataset is divided into multiple subsets, including test, easy chat, hard chat, safety, reasoning, etc., each with a specific number of examples and size. Additionally, a default configuration is provided, specifying the file paths for each dataset split.