MaximoLopezChenlo/OncoAgent-Clinical-266K

该数据集包含266,854个临床肿瘤学训练样本，专门为微调大语言模型在癌症诊断、治疗推荐和临床推理任务而精心整理。数据集由多个来源组成：约100,000个来自PubMed Central的真实临床案例展示（PMC-Patients），约50,000个基于证据的医学问答（PubMedQA），约90,000个思维链肿瘤学推理对（OncoCoT，合成数据），约20,000个来自85种以上癌症类型的结构化治疗协议（NCCN指南摘录），以及约6,854个欧洲临床实践指南（ESMO指南）。数据格式采用与trl.SFTTrainer兼容的聊天格式（JSONL），覆盖85种以上癌症类型，包括乳腺癌、肺癌、结直肠癌、前列腺癌等。数据集经过收集、过滤、去重、质量控制和格式化处理，仅用于研究和教育目的，不包含真实受保护健康信息（PHI）。This dataset contains 266,854 clinical oncology training samples curated for fine-tuning large language models on cancer diagnosis, treatment recommendation, and clinical reasoning tasks. It is composed of multiple sources: approximately 100,000 real clinical case presentations from PubMed Central (PMC-Patients), 50,000 evidence-based medical question answering (PubMedQA), 90,000 chain-of-thought oncology reasoning pairs (OncoCoT, synthetic), 20,000 structured treatment protocols from over 85 cancer types (NCCN Guideline Extracts), and 6,854 European clinical practice guidelines (ESMO Guidelines). The data follows a chat format (JSONL) compatible with trl.SFTTrainer, covering over 85 cancer types including breast, lung, colorectal, prostate, and others. The dataset underwent collection, filtering, deduplication, quality control, and formatting, intended for research and educational purposes only, with no real Protected Health Information (PHI).