Metraipolis › archive › dc_910994c84dd4

Still There, No Longer Seen: Exposing Compression-Induced Risk in Large Vision-Language Models — arXiv:2609.35002

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.35002
License (as archived)
CC0-1.0
Archived
2026-10-07T21:03:09.994Z
SHA-256 of the archived text
1ffcec240384e7c4c0e568e9ecfbc7500e6b1d10b49face1013ba2444ccc7e2e
Stable URL of this page
https://metraipolis.com/documento/dc_910994c84dd4

How to cite

Still There, No Longer Seen: Exposing Compression-Induced Risk in Large Vision-Language Models — arXiv:2609.35002 — arXiv (en), https://arxiv.org/abs/2609.35002 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_910994c84dd4 (Metraipolis archive dc_910994c84dd4, sha256 1ffcec240384e7c4c0e568e9ecfbc7500e6b1d10b49face1013ba2444ccc7e2e)
BibTeX
@misc{dc910994c84dd4,
  title = {Still There, No Longer Seen: Exposing Compression-Induced Risk in Large Vision-Language Models — arXiv:2609.35002},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.35002},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_910994c84dd4; sha256 1ffcec240384e7c4c0e568e9ecfbc7500e6b1d10b49face1013ba2444ccc7e2e},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.35002. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Visual token compression reduces the inference cost of Large Vision-Language Models (LVLMs). However, aggregate robustness measures do not reveal whether a particular adversarial failure is induced by compression or inherited from the underlying model. We define a compression-specific failure (CSF) as an adversarial input that remains correct under full-token inference but fails after compression, casting compression-induced risk as a paired failure attribution problem. Within a controlled diagnostic cohort, counterfactuals show that retained-set allocation causally changes compressed correctness and reveal a negative association between recovery and representation drift in displaced evidence. Motivated by these findings, we propose CIRA, a Compression-Induced Risk Attack for Large Vision-Language Models. Under a vision-encoder white-box setting, CIRA optimizes image perturbations through encoder-side objectives that manipulate token priorities across candidate compression budgets while preserving displaced evidence. CIRA uses no downstream questions or labels and requires no access to the language model, deployed compressor, or exact compression budget. Across 12 dataset-compressor settings evaluated at four budgets, CIRA achieves a mean CSFR of 20.35% while limiting full-token attack success to 6.92%, with similar behavior on additional LVLM families. A cross-view selection-stabilization defense substantially suppresses CIRA, although Adaptive CIRA partially restores its effectiveness. These results show that compression-specific failures persist under restricted access and support paired evaluation of full-token and compressed inference for attributing risk to visual-token compression.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-29 in cs.AI).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV) ; Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR).
Pagina del lavoro: https://arxiv.org/abs/2609.35002 — 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 2026-10-07T20:29:03.221Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).