General large models in the medical domain are prone to generating misinformation and require professional adaptation through domain data fine-tuning, RAG combined with authoritative medical literature, and other methods.