mssfj/openmathinstruct-2_formatted

OpenMathInstruct-2（格式化版）是一个指令调优数学语料库，源自OpenMathInstruct-2，并转换为简单的问答格式，包含明确的思维链痕迹。数据集包含13,972,791个示例（约23 GB磁盘空间），涵盖从小学到高级数学推理、代数、几何和文字问题。每个条目包含一个自然语言“问题”，一个以“<think>...</think>”推理部分开头并后跟“Final Answer:<value>”的“答案”，以及一个分类“category”标签。数据集以Parquet分片和JSONL转储形式提供，支持高效流式传输。

OpenMathInstruct-2 (formatted) is an instruction-tuning math corpus derived from OpenMathInstruct-2 and converted into a simple question–answer format with explicit chain-of-thought traces. The dataset contains 13,972,791 examples (about 23 GB on disk) covering grade-school to advanced math reasoning, algebra, geometry, and word problems. Each entry has a natural-language `question`, an `answer` that starts with a `<think>...</think>` reasoning section followed by `Final Answer:<value>`, and a categorical `category` tag. The dataset is packaged as a Parquet shard and a JSONL dump for efficient streaming.