TAUR-dev/raw__BC__math500_v3_distilled

该数据集包含了文档ID、文档内容（包括答案、难度等级、问题、解决方案、学科和唯一ID）、目标、参数（包括生成参数）、响应和过滤后的响应、文档哈希、提示哈希、目标哈希、精确匹配、提取的答案、行为提示、行为类型、行为跨度信息、新行为跨度计数和行为计数等字段。数据集被划分为训练集，共有399个示例，大小为6012732字节。

The dataset includes fields such as document ID, document content (including answer, difficulty level, problem, solution, subject, and unique ID), target, arguments (including generation parameters), responses and filtered responses, document hash, prompt hash, target hash, exact match, extracted answers, behavior prompt, behavior type, behavior span information, new behavior span count, and behavior count. The dataset is split into a training set with a total of 399 examples and a size of 6012732 bytes.