CMedINS
收藏资源简介:
CMedINS数据集是由中国科学院合肥物质科学研究院智能机器研究所与专业医生合作构建的医疗指令数据集,包含22万条真实医疗记录。该数据集涵盖了多个医疗部门的真实数据,经过严格的去标识化处理以保护患者隐私。数据集的内容包括医疗对话、医疗指令和一般能力数据,旨在通过监督微调和直接偏好优化方法提升模型在医疗领域的表现。该数据集的应用领域主要集中在医疗对话、诊断支持和患者咨询等方面,旨在解决医疗语言模型在专业领域知识不足的问题。
The CMedINS dataset is a medical instruction dataset constructed in collaboration with professional physicians by the Institute of Intelligent Machines, Hefei Institutes of Physical Science, Chinese Academy of Sciences, containing 220,000 real medical records. This dataset covers real data from multiple medical departments and has undergone strict de-identification processing to protect patient privacy. Its content includes medical dialogues, medical instructions and general capability data, aiming to improve the performance of models in the medical field through supervised fine-tuning and direct preference optimization methods. Its application scenarios mainly focus on medical dialogue, diagnostic support, patient consultation and other fields, aiming to address the problem of insufficient professional domain knowledge of medical language models.




