ReasonMed
收藏资源简介:
ReasonMed是一个大规模的医学推理数据集,包含37万个经过严格验证的示例,比之前的同类数据集大一个数量级。数据集由多个竞争性大型语言模型生成,并经过多阶段优化和验证流程,以确保推理路径的正确性和逻辑连贯性。数据集包含详细的推理过程和简洁的答案总结,有助于分析医学领域中的有效推理模式。ReasonMed旨在推动医学推理模型的发展,并解决医疗领域中知识密集型问答的挑战。
ReasonMed is a large-scale medical reasoning dataset consisting of 370,000 rigorously validated examples, which is an order of magnitude larger than prior comparable datasets. Generated by multiple competitive large language models, the dataset undergoes a multi-stage optimization and validation workflow to guarantee the correctness and logical coherence of its reasoning paths. It includes detailed reasoning processes and concise answer summaries, which facilitates the analysis of effective reasoning patterns in the medical field. ReasonMed aims to advance the development of medical reasoning models and address the challenges of knowledge-intensive question answering in the healthcare domain.

- 1通过阿里巴巴达摩院, 兰州大学基础医学院, 中国人民大学高瓴人工智能学院, 北京关键模型与智能治理实验室, 教育部下一代智能搜索与推荐工程研究中心, 湖畔实验室 · 2025年



