SubmissionNumber#=%=#20 FinalPaperTitle#=%=#1-2-3 Check: Enhancing Contextual Privacy in LLM via Multi-Agent Reasoning ShortPaperTitle#=%=# NumberOfPages#=%=#14 CopyrightSigned#=%=#Wenkai Li JobTitle#==# Organization#==#Carnegie Mellon University, 5000 Forbes Ave, Pittsburgh, PA 15213 Abstract#==#Addressing contextual privacy concerns remains challenging in interactive settings where large language models (LLMs) process information from multiple sources. Building on the theory of contextual integrity, we introduce a multi-agent framework that decomposes privacy reasoning into specialized subtasks—extraction, classification—reducing the information load on any single agent while enabling iterative validation and more reliable adherence to contextual privacy norms. Experiments on the ConfAIde benchmark with two LLMs (GPT-4, Llama3) demonstrate that our multi-agent system substantially reduces private information leakage (36\% reduction) while maintaining the fidelity of public content compared to a single-agent system, showing the promise of multi-agent frameworks towards contextual privacy with LLMs. Author{1}{Firstname}#=%=#Wenkai Author{1}{Lastname}#=%=#Li Author{1}{Username}#=%=#wenkai_li Author{1}{Email}#=%=#wenkai.kyle.li@gmail.com Author{1}{Affiliation}#=%=#Carnegie Mellon University Author{2}{Firstname}#=%=#Liwen Author{2}{Lastname}#=%=#Sun Author{2}{Username}#=%=#dominicslw Author{2}{Email}#=%=#liwens@andrew.cmu.edu Author{2}{Affiliation}#=%=#Carnegie Mellon University Author{3}{Firstname}#=%=#Zhenxiang Author{3}{Lastname}#=%=#Guan Author{3}{Email}#=%=#zhenxiag@andrew.cmu.edu Author{3}{Affiliation}#=%=#Carnegie Mellon University Author{4}{Firstname}#=%=#Xuhui Author{4}{Lastname}#=%=#Zhou Author{4}{Username}#=%=#xuhuiz Author{4}{Email}#=%=#xuhuiz@andrew.cmu.edu Author{4}{Affiliation}#=%=#Carnegie Mellon University Author{5}{Firstname}#=%=#Maarten Author{5}{Lastname}#=%=#Sap Author{5}{Username}#=%=#maartensap93 Author{5}{Email}#=%=#msap2@andrew.cmu.edu Author{5}{Affiliation}#=%=#Carnegie Mellon University ========== èéáğö