TAUR-dev/evals__syncotv5_with_bf16k__samples

该数据集包含文档ID、文档内容（包括答案、难度等级、问题、解决方案、学科和唯一标识符）、目标、参数（包括生成参数和子参数，如是否采样、最大生成token数、最大思考token数、温度、思考开始和结束标记等）、响应和过滤后的响应、文档哈希、提示哈希、目标哈希、精确匹配标志和提取的答案。数据集被划分为训练集，其中包含约500个示例，总大小约为14.6MB。

The dataset includes document ID, document content (including answer, level, problem, solution, subject, and unique identifier), target, arguments (including generation arguments and sub-arguments such as sampling, maximum generation tokens, maximum tokens for thinking, temperature, start and end of thinking, etc.), responses and filtered responses, document hash, prompt hash, target hash, exact match flag, and extracted answers. The dataset is split into a training set, which contains about 500 examples with a total size of approximately 14.6MB.