Metraipolis › archive › dc_eeea78b293d2

LEMUR: Latent Entropy-aware Multimodal Unlearning via Visual-anchored Reasoning Redirection — arXiv:2608.11691

Source: arXiv (en)
Original: https://arxiv.org/abs/2608.11691
License (as archived): CC0-1.0
Archived: 2026-10-19T05:30:58.794Z
SHA-256 of the archived text: 8e1712f40583d954e51b6ac9a08afd1fc257eccc47a887aad681f74d51d59077
Stable URL of this page: https://metraipolis.com/documento/dc_eeea78b293d2

How to cite

LEMUR: Latent Entropy-aware Multimodal Unlearning via Visual-anchored Reasoning Redirection — arXiv:2608.11691 — arXiv (en), https://arxiv.org/abs/2608.11691 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_eeea78b293d2 (Metraipolis archive dc_eeea78b293d2, sha256 8e1712f40583d954e51b6ac9a08afd1fc257eccc47a887aad681f74d51d59077)

BibTeX

@misc{dceeea78b293d2,
  title = {LEMUR: Latent Entropy-aware Multimodal Unlearning via Visual-anchored Reasoning Redirection — arXiv:2608.11691},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.11691},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_eeea78b293d2; sha256 8e1712f40583d954e51b6ac9a08afd1fc257eccc47a887aad681f74d51d59077},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.11691. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Reinforcement-learning (RL) post-training equips multimodal large reasoning models (MLRMs) with exploratory chains of thought (CoT), substantially improving visual reasoning. However, we find that this capability introduces a distinct privacy vulnerability: even when a sensitive fact is successfully unlearned from the final answer, the model may still reproduce it in its reasoning trace. This leakage is substantially more pronounced in natively RL-trained MLRMs than in their non -reasoning base models, revealing a privacy risk that existing unlearning methods are not designed to address. We show that RL-induced exploration leaves sensitive content with a distinctive token-level entropy signature that is largely absent from base models. Based on this observation, we propose LEMUR, a fully training-free, inference-time unlearning framework for natively RL-trained multimodal models. LEMUR uses entropy dynamics as a control signal to identify when sensitive reasoning begins and when sanitization should stop. During this interval, it redirects the reasoning trajectory through entropy-modulated visual-anchor latent injection, replacing committed tokens with sanitized, probability-weighted embeddings re-grounded in the input image. Across diverse MLRMs, LEMUR consistently outperforms existing unlearning met hods in suppressing both reasoning-trace and answer leakage, while better preserving non-sensitive utility and output fluency. These results demonstrate that RL-induced entropy dynamics provide a distinctive signal for privacy leakage and that exploiting this signal enables effective training-free unlearning for reasoning-capable multimodal models.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-13 in cs.CL).
Categorie arXiv: Machine Learning (cs.LG) ; Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2608.11691 — 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 2026-10-19T05:15:39.249Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-08-13?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).