Researchers have introduced BusMA, a novel communication framework designed for multi-agent systems that aims to enhance agent autonomy and efficiency. Unlike existing Hierarchical Manager-Worker (HMW) or Router-based Message Passing (RMP) structures, BusMA utilizes a shared bus architecture allowing any agent to directly address others. This system includes components for agent registration, message routing, and shared memory management, enabling workers to post messages with specific intents like discussion, challenge, or guidance. Experiments using frontier LLMs across 13 tasks demonstrated that BusMA consistently outperforms current HMW and RMP methods in visual reasoning, mathematical reasoning, and knowledge retrieval. AI
IMPACT Enhances agent autonomy and efficiency in complex tasks, potentially improving performance in areas like reasoning and knowledge retrieval.
RANK_REASON The cluster contains a research paper detailing a new framework for multi-agent systems. [lever_c_demoted from research: ic=1 ai=1.0]
- Chair agent
- Hierarchical Manager-Worker (HMW)
- Multi-Agent (MA) systems
- Router-based Message Passing (RMP)
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
