Foundations of Medical LLMs
A Chinese-language, monthly-updated textbook on medical large language models, covering foundational LLM technique through vertical training for healthcare, clinical applications, and the field's ethical and safety challenges.
Foundations of Medical LLMs is an open, continuously updated textbook that explains large language models for a healthcare audience, from general LLM fundamentals to medical-specific applications. Its first edition is organized into five parts: the evolution from general AI to medical LLMs; technical foundations such as the Transformer architecture, pretraining and fine-tuning, and prompt engineering; the move from general-domain to vertical medical training, including medical data preparation, frontier techniques and evaluation benchmarks; frontier clinical and research applications such as clinical decision support, automated medical documentation and drug discovery; and challenges around hallucination, data privacy, medical ethics and bias, and the outlook for multimodal medical AI.
Each chapter is intended to carry an accompanying, continuously tracked list of relevant papers so the material stays current with new research.
It is written in Chinese and updated monthly by its author team, who invite corrections and suggestions through GitHub issues.