TAUR-dev/evals__bf_16k_lora_syncot_v1__samples

这个数据集包含了文档ID、文档内容（答案、难度级别、问题、解决方案、主题和唯一ID）、目标、参数（包括生成参数和它们的属性，如是否采样、最大生成标记数、最大思考标记数、温度、思考开始和结束标记、直到标记）、响应、过滤后的响应、文档哈希、提示哈希、目标哈希、精确匹配和提取的答案等字段。训练集包含500个示例，数据集总大小为28,680,413字节。

The dataset includes fields such as document ID, document content (answer, level, problem, solution, subject, and unique ID), target, arguments (including generation arguments and their properties like sampling, maximum generation tokens, maximum tokens for thinking, temperature, start and end of thinking, and until token), responses, filtered responses, document hash, prompt hash, target hash, exact match, and extracted answers. The training set contains 500 examples, with the total dataset size being 28,680,413 bytes.