Metraipolis › archive › dc_dc7c10079f47

Where Identity Lives: Localized, Retain-Free Identity Unlearning in Multimodal Large Language Models — arXiv:2608.30649

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.30649
License (as archived)
CC0-1.0
Archived
2026-10-17T08:07:40.160Z
SHA-256 of the archived text
c1c5888a8b31f4c14ce6abc9c0628719da3e77242517f3f005ee751d15d292ee
Stable URL of this page
https://metraipolis.com/documento/dc_dc7c10079f47

How to cite

Where Identity Lives: Localized, Retain-Free Identity Unlearning in Multimodal Large Language Models — arXiv:2608.30649 — arXiv (en), https://arxiv.org/abs/2608.30649 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_dc7c10079f47 (Metraipolis archive dc_dc7c10079f47, sha256 c1c5888a8b31f4c14ce6abc9c0628719da3e77242517f3f005ee751d15d292ee)
BibTeX
@misc{dcdc7c10079f47,
  title = {Where Identity Lives: Localized, Retain-Free Identity Unlearning in Multimodal Large Language Models — arXiv:2608.30649},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.30649},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_dc7c10079f47; sha256 c1c5888a8b31f4c14ce6abc9c0628719da3e77242517f3f005ee751d15d292ee},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.30649. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Removing a specific individual's information from multimodal large language models (MLLMs) is often needed after deployment, but existing methods rely on a retain set, which is hardest to obtain at that point, and rebuilding it recreates the privacy exposure that unlearning aims to remove. Forgetting from the forget set alone instead damages the shared visual-language computation, harming perception. We cast retain-free unlearning as a localization problem: causal tracing, weight transplant, and Fisher overlap all point to early-to-mid decoder MLPs as the layers where identity information is stored and, unlike other module families, can be modified without substantially disrupting vision. We turn this into Pathway-Aware Visual-attribute Anchoring (PAVA), which confines updates to these layers and pairs a forget loss with a visual-attribute anchor that preserves image-grounded behavior by distilling the model's own pre-unlearning answers from the forget images alone. On MLLMU-Bench and ReMem, PAVA gives the strongest forget-retain trade-off among forget-set-only methods and remains competitive with retain-based baselines.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-01 in cs.CL).
Categorie arXiv: Computation and Language (cs.CL) ; Computer Vision and Pattern Recognition (cs.CV).
Pagina del lavoro: https://arxiv.org/abs/2608.30649 — 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 2026-10-17T07:47:52.838Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-09-01?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).