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 2610.19458cs.AI
收藏

PFedMA:联邦多智能体编排演进中,谁应加入下一步?

PFedMA: Who Should Join Next as Federated Multi-Agent Orchestration Evolves?

Tianjin University of Technology(天津工业大学);Inner Mongolia University(内蒙古大学);Institute of automation, Chinese academy of science(中国科学院自动化研究所)

Shuai Li, Shengyuan Liu, Fan Qi, Xiaoshan Yang

AI总结:

针对联邦多智能体协作的智能体选择与隐私规则适配难题,提出PFedMA框架,通过维护历史状态、双向隐私规则演进等提升任务性能、降低隐私泄露并增强鲁棒性。

摘要:

联邦多智能体系统(FedMAS)为基于大语言模型(LLM)的智能体提供了一种协作推理的有前景范式,同时保持原始客户端数据去中心化。然而,可靠的多轮协作仍具挑战性:智能体选择需考虑不断变化的任务需求和智能体的历史可靠性,而隐私规则必须适应不断变化的协作情境并满足客户端的个性化需求。为解决这些挑战,本文提出PFedMA,这是一种用于异构隐私要求下反馈驱动多轮协作的联邦多智能体框架。在多轮过程中,PFedMA维护历史能力与可靠性状态以指导智能体选择,在聚合前评估任务响应,并根据积累的协作反馈优化后续查询。PFedMA支持跨异构客户端的本地到全局、全局到本地的双向隐私规则演进;客户端泛化本地更新的隐私规则并选择性纳入全局维护的规则,从而在保留个性化约束的同时促进跨客户端隐私规则演进。在推理、隐私敏感和对抗情境下的大量实验表明,PFedMA可提升任务性能、减少测得的隐私泄露并增强系统鲁棒性。

英文摘要:

Federated Multi-Agent Systems (FedMAS) offer a promising paradigm for collaborative reasoning among LLM-based agents while keeping raw client data decentralized. However, reliable multi-round collaboration remains challenging. Agent selection must take into consideration evolving task requirements and the agent's historical reliability, while privacy rules must adapt to changing collaboration contexts and meet the client's personalized needs. To address these challenges, we propose PFedMA, a federated multi-agent framework for feedback driven multi-round collaboration under heterogeneous privacy requirements. Across rounds, PFedMA maintains historical capability and reliability states to guide agent selection, evaluates task responses before aggregation, and refines subsequent queries according to accumulated collaboration feedback. PFedMA enables bidirectional local to global and global to local private rule evolution across heterogeneous clients. Clients generalize locally updated privacy rules and selectively incorporate globally maintained rules, thereby facilitating cross-client privacy rule evolution while preserving personalized constraints. Extensive experiments across reasoning, privacy sensitive, and adversarial settings show that PFedMA improves task performance, reduces measured privacy leakage, and improves system robustness.

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。

↑