Salvato in:
Dettagli Bibliografici
Autori principali: Cheng, Yuhan, Ye, Hancheng, Li, Hai Helen, Sun, Jingwei, Chen, Yiran
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:https://arxiv.org/abs/2602.13840
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866912906021437440
author Cheng, Yuhan
Ye, Hancheng
Li, Hai Helen
Sun, Jingwei
Chen, Yiran
author_facet Cheng, Yuhan
Ye, Hancheng
Li, Hai Helen
Sun, Jingwei
Chen, Yiran
contents Large language model (LLM) agents are increasingly deployed in personalized tasks involving sensitive, context-dependent information, where privacy violations may arise in agents' action due to the implicitness of contextual privacy. Existing approaches rely on external, inference-time interventions which are brittle, scenario-specific, and may expand the privacy attack surface. We propose PrivAct, a contextual privacy-aware multi-agent learning framework that internalizes contextual privacy preservation directly into models' generation behavior for privacy-compliant agentic actions. By embedding privacy preferences into each agent, PrivAct enhances system-wide contextual integrity while achieving a more favorable privacy-helpfulness tradeoff. Experiments across multiple LLM backbones and benchmarks demonstrate consistent improvements in contextual privacy preservation, reducing leakage rates by up to 12.32% while maintaining comparable helpfulness, as well as zero-shot generalization and robustness across diverse multi-agent topologies. Code is available at https://github.com/chengyh23/PrivAct.
format Preprint
id arxiv_https___arxiv_org_abs_2602_13840
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle PrivAct: Internalizing Contextual Privacy Preservation via Multi-Agent Preference Training
Cheng, Yuhan
Ye, Hancheng
Li, Hai Helen
Sun, Jingwei
Chen, Yiran
Computation and Language
Large language model (LLM) agents are increasingly deployed in personalized tasks involving sensitive, context-dependent information, where privacy violations may arise in agents' action due to the implicitness of contextual privacy. Existing approaches rely on external, inference-time interventions which are brittle, scenario-specific, and may expand the privacy attack surface. We propose PrivAct, a contextual privacy-aware multi-agent learning framework that internalizes contextual privacy preservation directly into models' generation behavior for privacy-compliant agentic actions. By embedding privacy preferences into each agent, PrivAct enhances system-wide contextual integrity while achieving a more favorable privacy-helpfulness tradeoff. Experiments across multiple LLM backbones and benchmarks demonstrate consistent improvements in contextual privacy preservation, reducing leakage rates by up to 12.32% while maintaining comparable helpfulness, as well as zero-shot generalization and robustness across diverse multi-agent topologies. Code is available at https://github.com/chengyh23/PrivAct.
title PrivAct: Internalizing Contextual Privacy Preservation via Multi-Agent Preference Training
topic Computation and Language
url https://arxiv.org/abs/2602.13840