Abstract
Large Language Model (LLM)-driven interactive systems currently showpotential promise in healthcare domains. Despite their remarkable capabilities,LLMs typically lack personalized recommendations and diagnosis analysis insophisticated medical applications, causing hallucinations and performancebottlenecks. To address these challenges, this paper proposes MedAide, anLLM-based omni medical multi-agent collaboration framework for specializedhealthcare services. Specifically, MedAide first performs query rewritingthrough retrieval-augmented generation to accomplish accurate medical intentunderstanding. Immediately, we devise a contextual encoder to obtain intentprototype embeddings, which are used to recognize fine-grained intents bysimilarity matching. According to the intent relevance, the activated agentscollaborate effectively to provide integrated decision analysis. Extensiveexperiments are conducted on four medical benchmarks with composite intents.Experimental results from automated metrics and expert doctor evaluations showthat MedAide outperforms current LLMs and improves their medical proficiencyand strategic reasoning.