库帕思高质量消防手册思维链（Chain-of-Thought）数据集

本数据集面向消防领域大模型及智能问答系统研发团队，旨在解决通用大模型在消防实务中存在的流程描述不准确、操作要点缺失、处置逻辑不清晰等问题。以《中国消防手册》为基础，通过大模型生成并经专业核验的思维链推理样本，完整呈现“问题识别→流程分析→操作要点→处置建议”的专业推理路径，覆盖火灾预防、灭火救援、装备使用、应急管理等核心章节。

通过思维链训练，可提升模型在消防实务操作指导、应急处置流程、装备使用方法等方面的推理能力与决策透明度。与传统标准问答数据集不同，本数据集将每个正确答案背后的专业判断逻辑完整展开，使模型不仅“知道做什么”，更能“理解为什么这么做”。

This dataset is targeted at R&D teams of large language models (LLMs) and intelligent question answering systems in the fire protection domain, aiming to address issues existing in general LLMs when applied to fire protection practice, such as inaccurate process descriptions, missing operational key points, and unclear disposal logic. Based on the *China Fire Protection Manual*, this dataset provides Chain-of-Thought (CoT) reasoning samples generated by LLMs and verified by professionals, which fully present the professional reasoning path of "Problem Identification → Process Analysis → Operational Key Points → Disposal Suggestions", covering core chapters including fire prevention, fire fighting and rescue, equipment operation, emergency management and other related fields. Through CoT-based training, the model's reasoning ability and decision-making transparency in aspects such as fire protection practical operation guidance, emergency disposal procedures, and equipment operation methods can be enhanced. Unlike traditional standard question-answer datasets, this dataset fully elaborates the professional judgment logic behind each correct answer, enabling the model to not only "know what to do" but also "understand why it is done this way".

搜集汇总

背景与挑战

背景概述

该数据集面向消防领域大模型研发，基于《中国消防手册》构建了经过专业核验的思维链推理样本，覆盖火灾预防、灭火救援等核心内容。它旨在通过呈现“问题识别→处置建议”的完整推理路径，提升模型在消防实务中的决策透明度和准确性。