mssfj/openmathinstruct-2_structured-1000

OpenMathInstruct-2 Structured (CoT)是一个用于数学任务监督微调的数据集，旨在通过结构化的思维链（Chain-of-Thought）稳定生成长步骤的数学问题解答。数据集包含1,446条英语记录，每条记录包含问题文本（question）、带有结构化CoT和最终答案的文本（answer）以及任务类别（category）。结构化CoT包括<analyze>、<plan>、<verify>等标签，帮助模型逐步推理并生成最终答案。数据集由gpt-oss-120b模型自动生成，可能存在计算或描述错误，建议在使用前进行验证和清理。

OpenMathInstruct-2 Structured (CoT) is a dataset designed for supervised fine-tuning of mathematical tasks, aiming to generate long-step math problem solutions stably using structured Chain-of-Thought (CoT). The dataset contains 1,446 English records, each consisting of a question text (question), an answer text with structured CoT and final answer (answer), and a task category (category). The structured CoT includes tags such as <analyze>, <plan>, and <verify> to guide the model through step-by-step reasoning. The dataset is auto-generated by the gpt-oss-120b model and may contain calculation or descriptive errors, so verification and cleaning are recommended before use.