Paper proposes action-state communication for multi-agent LLMs

Arxiv CS.AI (top papers)

The paper proposes action-state communication for multi-agent LLM systems, where agents exchange structured action-state messages instead of free-form natural language. This approach aims to reduce redundant information and improve the efficiency of inter-agent communication.

multi-agent-systemsllm

Arxiv CS.AI (top papers)·Chen HuangYuhao WuWenxuan Zhang

6 days ago