发表机构
The Hong Kong Polytechnic University; National University of Singapore; The University of Tokyo; Hong Kong University of Science and Technology; Tengen AI（香港理工大学; 新加坡国立大学; 东京大学; 香港科技大学; 腾根人工智能）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本文提出联邦智能体优化（FAO），通过受控信息交换实现分布式LLM智能体协作提升，平衡效用、隐私与通信成本，并探索私有经验转化为可迁移能力的途径。
AI 中文摘要
大型语言模型（LLM）智能体日益在私有环境中运行，并从任务执行、工具使用、反馈和本地知识中积累宝贵经验。然而，这些经验分布在不同的组织中，由于隐私和专有性约束，无法直接共享。传统的联邦学习不足以应对这一场景，因为智能体的能力不仅限于模型参数，还扩展至记忆、工具、奖励、技能和结构化知识。在本文中，我们提出了联邦智能体优化（Federated Agent Optimization, FAO），研究分布式智能体如何通过受控的信息交换进行协作改进，同时将原始数据、完整轨迹和私有知识保留在本地。我们将FAO定义为一个多目标优化问题，在智能体效用、隐私泄露和通信成本之间进行平衡，并组织其在策略、记忆、工具使用、奖励以及结构化知识和技能方面的优化空间。我们进一步刻画了私有经验如何被抽象、保护、聚合和适应为可迁移的能力，提供了一个统一视角，说明智能体如何在无需直接经验共享的情况下相互受益。最后，我们识别了FAO的关键挑战，并概述了未来研究朝着可信联邦智能体系统发展的几个有前景的方向。
英文摘要
Large language model (LLM) agents increasingly operate in private environments and accumulate valuable experience from task execution, tool use, feedback, and local knowledge. Yet such experience is distributed across organizations and cannot be directly shared because of privacy and proprietary constraints. Conventional federated learning is insufficient for this setting, as agent capabilities extend beyond model parameters to memory, tools, rewards, skills, and structured knowledge. In this paper, we formulate \textbf{Federated Agent Optimization (FAO)}, which studies how distributed agents can collaboratively improve through controlled information exchange while keeping raw data, complete trajectories, and private knowledge local. We define FAO as a multi-objective problem balancing agent utility, privacy leakage, and communication cost, and organize its optimization space across policy, memory, tool use, reward, and structured knowledge and skills. We further characterize how private experience can be abstracted, protected, aggregated, and adapted into transferable capabilities, providing a unified view of how agents can benefit from one another without direct experience sharing. Finally, we identify the key challenges of FAO and outline several promising directions for future research toward trustworthy federated agent systems.
