大型语言模型(LLM)代理已经 emerged 作为复杂任务的强大工具,但它们适应个体用户的能力仍然从根本上受到限制。我们主张这种限制源于一个关键的架构混淆:当前系统将记忆、学习和个性化视为统一能力,而不是需要不同基础设施、不同时间尺度和需要独立优化的三个独立机制。我们提出了MAPLE(Memory-Adaptive Personalized LEarning),一种原则性的分解,其中记忆处理存储和检索基础设施;学习从积累的交互中异步提取智能;个性化在有限上下文预算内实时应用所学知识。每个组件都作为专门的子代理运行,具有专门的工具和明确的接口。在MAPLE-Personas基准上的实验评估表明,我们的分解在与无状态基线相比时,个性化得分提高了14.6%(p < 0.01,Cohen's d = 0.95),并使特征纳入率从45%提高到75%——使代理真正学会并适应。
英文摘要:
Large language model (LLM) agents have emerged as powerful tools for complex tasks, yet their ability to adapt to individual users remains fundamentally limited. We argue this limitation stems from a critical architectural conflation: current systems treat memory, learning, and personalization as a unified capability rather than three distinct mechanisms requiring different infrastructure, operating on different timescales, and benefiting from independent optimization. We propose MAPLE (Memory-Adaptive Personalized LEarning), a principled decomposition where Memory handles storage and retrieval infrastructure; Learning extracts intelligence from accumulated interactions asynchronously; and Personalization applies learned knowledge in real-time within finite context budgets. Each component operates as a dedicated sub-agent with specialized tooling and well-defined interfaces. Experimental evaluation on the MAPLE-Personas benchmark demonstrates that our decomposition achieves a 14.6% improvement in personalization score compared to a stateless baseline (p < 0.01, Cohen's d = 0.95) and increases trait incorporation rate from 45% to 75% -- enabling agents that genuinely learn and adapt.
PersonaMem-v2: Towards Personalized Intelligence via Learning Implicit User Personas and Agentic Memory
University of Pennsylvania(宾夕法尼亚大学);Massachusetts Institute of Technology(麻省理工学院);University of Washington(华盛顿大学);University of California Santa Barbara(加州大学圣巴巴拉分校);Meta;University of Southern California(南加州大学);University of North Carolina at Chapel Hill(北卡罗来纳大学教堂山分校);Microsoft Corporation(微软公司)
AgenticShop: Benchmarking Agentic Product Curation for Personalized Web Shopping
Department of Artificial Intelligence Yonsei University Seoul Republic of Korea(人工智能系 首尔国立庆尚大学 首尔 大韩民国);ParamitaAI Seoul Republic of Korea(ParamitaAI 首尔 大韩民国);Yonsei University(首尔国立庆尚大学);ParamitaAI
Towards trustworthy agentic AI: a comprehensive survey of safety, robustness, privacy, and system security
Faculty of Engineering, Department of Computer Science and Engineering, The Chinese University of Hong Kong(香港中文大学工程学院、计算机科学与工程系);Artificial Intelligence Innovation and Incubation Institute, Fudan University(复旦大学人工智能创新与孵化院);Shanghai Academy of AI for Science(上海人工智能科学研究院)
FinAgent: An Agentic AI Framework Integrating Personal Finance and Nutrition Planning
Faculty of Computer and Information System, Islamic University of Madinah, Saudi Arabia(麦加伊斯兰大学计算机与信息学院);Department of Computer Science, Effat College of Engineering, Effat University, Saudi Arabia(Effat大学工程学院计算机科学系);Arab Open University, Bahrain(巴林开放大学)
An Agentic Operationalization of DISARM for FIMI Investigation on Social Media
Center of Research Acquisition National Institute of Cyber Security Taipei City, Taiwan;Department of Mathematics Royal Military Academy Brussels, Belgium;Research Division OpenMinds Ltd. Kyiv, Ukraine;Department of Sustainable Communication Technologies SINTEF Oslo, Norway (Corresponding Author)
Cognitive Edge Computing: A Comprehensive Survey on Optimizing Large Models and AI Agents for Pervasive Deployment
Beijing Normal-Hong Kong Baptist University(北京师范大学-香港浸会大学联合国际学院);Institute of Artificial Intelligence and Future Networks, Beijing Normal University (Zhuhai)(北京师范大学(珠海)人工智能与未来网络研究院);The Hong Kong Polytechnic University(香港理工大学)
Beyond Similarity: Trustworthy Memory Search for Personal AI Agents
University of Science and Technology of China(中国科学技术大学);Tsinghua University(清华大学);National University of Singapore(新加坡国立大学);University of California, Berkeley(加州大学伯克利分校)
A Survey of Agent Memory in the Second Half: Towards Self-Evolving and Long-Horizon Agents
UIC(伊利诺伊大学香槟分校);IIT(伊利诺伊理工学院);UIUC(伊利诺伊大学厄巴纳-香槟分校);UW–Madison(威斯康星大学麦迪逊分校);ASU(亚利桑那州立大学);Emory(埃默里大学);Northwestern(西北大学);NTU(国立台湾大学);UCF(佛罗里达大学);Rutgers(新泽西罗格斯大学);Cambridge(剑桥大学);Harvard(哈佛大学);UTokyo(东京大学);UCSD(加州大学圣地亚哥分校);UCSC(加州大学圣塔克鲁斯分校);TAMU(德克萨斯大学奥斯汀分校);UCSB(加州大学圣塔芭芭拉分校);MBZUAI(马克斯·普朗克人工智能研究所);McGill(麦吉尔大学);UCLA(加州大学洛杉矶分校);Stanford(斯坦福大学);Salesforce;Google(谷歌);Meta;Roblox;Cisco(思科);Capital One
Agentic AI Framework for Individuals with Disabilities and Neurodivergence: A Multi-Agent System for Healthy Eating, Daily Routines, and Inclusive Well-Being
Hidden in Memory: Sleeper Memory Poisoning in LLM Agents
SPAR;ELLIS Institute Tübingen(图宾根ELLIS研究所);MPI for Intelligent Systems(智能系统马克斯·普朗克研究所);Tübingen AI Center(图宾根人工智能中心);APTA;CISPA Helmholtz Center for Information Security(信息安全海德堡中心)