qingy2024/FineQwQ-142k

该数据集包含提示（prompt）和响应（response）对，分为10k、25k、50k、100k和142k五种不同大小的数据集。数据来源于三个不同的原始资源，分别占总行数的不同比例，专注于不同类型的问题或推理任务。具体来源包括：powerinfer/qwq-500k（仅保留编码问题以避免重复，占35.84%）、qwq-longcot-verified（经过验证的数学问题，占45.14%）、amphora-magpie（多样化的通用推理，占19.02%）。

The dataset includes pairs of prompts and responses, split into sizes of 10k, 25k, 50k, 100k, and 142k. It is sourced from three different origins, each contributing a different percentage of the total rows and focusing on different types of problems or reasoning tasks: powerinfer/qwq-500k (coding problems to avoid overlap, 35.84%), qwq-longcot-verified (verified math problems, 45.14%), amphora-magpie (diverse general purpose reasoning, 19.02%).