nuhmanpk/firstaid-treatment-instruct

FirstAid-Treatment IT是一个用于训练领域专用语言模型的指令微调和问答数据集，专注于急救、紧急响应、程序性医疗支持和安全指导。数据集包含基于源对齐的急救参考材料和紧急响应文档构建的对话式训练示例。适用于监督微调(SFT)、问答(QA)、检索增强生成(RAG)、基于医疗助理训练和安全导向的指令遵循模型。数据集覆盖多种急救主题，如创伤管理、出血控制、中毒响应等。每个示例遵循聊天式指令微调格式，包含消息列表、来源和主题。数据集创建过程包括源数据摄取、文档分块、主题提取、指令合成、基于答案生成、质量过滤和对话格式化。

FirstAid-Treatment IT is an instruction tuning and question-answering dataset for training domain-specialized language models on first aid, emergency response, procedural medical support, and safety guidance. The dataset consists of grounded conversational training examples built from source-aligned first-aid reference material and emergency response documentation. Designed for Supervised Fine-Tuning (SFT), Question Answering (QA), Retrieval-Augmented Generation (RAG), grounded medical assistant training, and safety-focused instruction-following models. Covers diverse first aid topics such as trauma management, bleeding control, poisoning response, etc. Each example follows chat-style instruction tuning format with messages list, source, and topic. Dataset creation involves source ingestion, document chunking, topic extraction, instruction synthesis, grounded answer generation, quality filtering, and conversational formatting.