Researchers have developed SMART-Med, a novel multi-agent system for medical reasoning that challenges the assumption of constant collaboration. Instead, SMART-Med adaptively decides when to answer queries directly with a single agent and when to involve a subset of expert agents, based on an uncertainty-driven routing mechanism. This selective collaboration significantly reduces inference costs, including a 3.8x decrease in inference time and a 72% reduction in token usage, while maintaining competitive performance on medical visual question answering benchmarks. AI
IMPACT Reduces computational costs for AI-driven medical analysis, potentially enabling wider deployment of advanced reasoning systems.
RANK_REASON Research paper detailing a new method for multi-agent medical reasoning. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
