Med-ChatGLM

Why we included this project

Chinese-language medical assistants are still a young space, and this project gives teams a concrete view of how to adapt a general bilingual model to clinical questions. It fine-tunes ChatGLM-6B on an instruction set assembled from a medical knowledge graph and GPT-3.5-generated Q&A pairs, with coverage spanning diseases, drugs, and lab indicators. Alongside the trained checkpoint, the repo documents how the data was built and the tuning script used, so you can re-run the pipeline on your own clinical data. The authors note the dataset is still being refined, which matters if you plan to build on it. Treat the release as a working reference for domain-specific instruction tuning rather than a finished product you can serve patients with.