Masahiro Kaneko, Hiromi Arai, Timothy Baldwin · 2026 · arXiv
DOI: 10.48550/arXiv.2610.28315 · open-access PDF
Los agentes del modelo de lenguaje grande (LLM) utilizan cada vez más la memoria persistente para personalizar las respuestas, y los atributos personales en esta memoria suponen un riesgo de privacidad si se exponen. Nos preguntamos cómo la memoria de consulta durante el diálogo cambia lo que los usuarios van a decir, y por lo tanto lo que la memoria viene a revelar. En las simulaciones de múltiples sesiones, un usuario simulado con 24 atributos personales habla con un agente de apoyo que consulta o no la memoria, y en ambos ajustes los atributos se deducen del registro acumulado completo.
Machine translation (Meta M2M100); the English original follows.
Large language model (LLM) agents increasingly use persistent memory to personalize responses, and personal attributes in this memory pose a privacy risk if exposed. We ask how consulting memory during dialogue changes what users go on to say, and thus what the memory comes to reveal. In multi-session simulations, a simulated user with 24 personal attributes talks with a support agent that does or does not consult memory, and in both settings attributes are inferred from the full accumulated record. Across four support models, memory raises response quality but also empirical leakage, the shar
Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese
GET https://store.agentexchange.work/gw/papers/verify?title=...&year=... (0.003 USD), fetch this record as JSON with /gw/papers/paper?id=2610.28315, or list what it cites with /gw/papers/references?id=.... Pay per call in USDC (x402) or with a 19 USD card key for 1,000 calls. Catalogue · llms.txtSources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange