arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2605.29612cs.MAcs.CL
收藏

CONCAT: 基于共识与置信驱动的即席团队协作以实现高效的基于LLM的多智能体系统

CONCAT: Consensus- and Confidence-Driven Ad Hoc Teaming for Efficient LLM-Based Multi-Agent Systems

Southeast University(东南大学);Huawei Technologies Ltd(华为技术有限公司)

Ziyang Ma, Dingyi Zhang, Sichu Liang, Jiajia Chu, Pengfei Xia, Hui Zang, Deyu Zhou

AI总结:

提出一种无需训练的共识与置信驱动即席团队协作框架CONCAT,通过聚类初始答案、选择高置信领导者并基于心智理论预测协作收益来动态组织多智能体交互,显著提升效率并降低延迟。

摘要:

尽管基于大型语言模型的多智能体系统在解决复杂任务和实现比单智能体系统更高的性能方面显示出能力,但由于智能体之间的密集通信,它们导致了巨大的计算开销。先前的研究致力于训练稀疏多智能体图或微调规划器以更好地编排工作流程。然而,这些额外的训练过程引入了计算成本,并将多智能体系统限制在特定领域,从而损害了其泛化能力。在本文中,我们提出了CONCAT,一种基于共识和置信驱动的即席团队协作的无训练多智能体协作框架,以高效组织智能体交互。具体来说,智能体根据其初始答案进行聚类,并根据智能体的置信度选择每个聚类的领导者。然后,基于心智理论设计启发式函数,根据领导者的答案和置信度预测每两个领导者之间的协作收益。最后,在根据预测收益驱逐一定比例的通信后,组织一个即席多智能体网络。在三个LLM和三个基准上的实验表明,CONCAT比LLM-Debate实现了高达2.02倍的效率(准确率/延迟比),并优于诸如AgentDropout等训练感知方法,同时在Qwen2.5-14B-Instruct上将平均延迟降低了50.1%,且无需任何任务特定训练。

英文摘要:

Although large language model (LLM) based multi-agent systems (MAS) show their capability to solve complex tasks and achieve higher performance over single agent systems, they lead to huge computational overheads because of heavy communication between agents. Previous research has made efforts to train a sparse multi-agent graph or fine-tune a planner to orchestrate the workflow better. However, such extra training processes introduce computational costs and limit MAS to specific domains, therefore compromising their generalizability. In this paper, we propose CONCAT, a training-free multi-agent collaboration framework based on CONsensus and Confidence-driven Ad hoc Teaming to efficiently organize agent interactions. Specifically, agents are clustered based on their initial answers, and leaders of each cluster are selected based on the agents' confidence. Then, a heuristic function based on the Theory of Mind is designed to predict the collaboration benefits between every two leaders according to their answers and confidence. Finally, an ad hoc multi-agent network is organized after evicting a percentage of communications based on the predicted benefits. Experiments across three LLMs and three benchmarks show that CONCAT achieves up to 2.02x higher efficiency (accuracy/latency ratio) than LLM-Debate and outperforms training-aware methods such as AgentDropout, while reducing average latency by 50.1% on Qwen2.5-14B-Instruct, without any task-specific training.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑