Metraipolis › archive › dc_af4903809d7a

CAPTURE: Disentangling Preference Drift from Memory Poisoning in Personalized LLM Agents — arXiv:2609.02265

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.02265
License (as archived)
CC0-1.0
Archived
2026-10-17T04:59:28.362Z
SHA-256 of the archived text
fdfbc02b11cc51c09a6aeeee7745854c47c7829e1c399150e9a16dd0dfd13a80
Stable URL of this page
https://metraipolis.com/documento/dc_af4903809d7a

How to cite

CAPTURE: Disentangling Preference Drift from Memory Poisoning in Personalized LLM Agents — arXiv:2609.02265 — arXiv (en), https://arxiv.org/abs/2609.02265 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_af4903809d7a (Metraipolis archive dc_af4903809d7a, sha256 fdfbc02b11cc51c09a6aeeee7745854c47c7829e1c399150e9a16dd0dfd13a80)
BibTeX
@misc{dcaf4903809d7a,
  title = {CAPTURE: Disentangling Preference Drift from Memory Poisoning in Personalized LLM Agents — arXiv:2609.02265},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.02265},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_af4903809d7a; sha256 fdfbc02b11cc51c09a6aeeee7745854c47c7829e1c399150e9a16dd0dfd13a80},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.02265. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Personalized language agents use persistent memory to adapt to users over time, but the same mechanism creates an attack surface. When new information conflicts with stored preferences, an agent must distinguish genuine preference drift from temporary context shifts, ambiguity, or adversarial memory poisoning. We formulate this problem as a continuous-time partially observable decision process over a latent user state and show why rules based only on recency and provenance are insufficient. CAPTURE addresses this ambiguity with a neural differential-equation belief tracker, a multi-timescale memory ledger, uncertainty-triggered clarification, and counterfactual auditing of cited memories. On 480 held-out episodes from 96 users, CAPTURE achieves a 71.5% win rate, compared with 69.3% for an identically supervised baseline and 66.1% for the strongest heuristic baseline. It limits fixed-policy poisoning success to 11.5% while accepting 83.5% of genuine preference updates. Under an adaptive attacker with access to the released weights, attack success rises to 24.7%, exposing a real adaptation-security tradeoff. We further evaluate the frozen system zero-shot on an independently constructed benchmark and replay longitudinal interaction histories from 40 users collected over two to three weeks. These results suggest that modeling preference authenticity explicitly can improve both personalization and robustness in memory-augmented LLM agents.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-03 in cs.LG).
Categorie arXiv: Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2609.02265 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-17T04:09:43.320Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-09-03?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).