Metraipolis › archive › dc_38583a7bf73e
Understanding and Mitigating Token-Pruning-Induced Vulnerabilities in VLMs — arXiv:2610.09703
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.09703
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-08T11:44:24.816Z
- SHA-256 of the archived text
60feb462313238c45863fd513b7bf2df50d7a89329fc381a29f002c874086052- Stable URL of this page
https://metraipolis.com/documento/dc_38583a7bf73e
How to cite
Understanding and Mitigating Token-Pruning-Induced Vulnerabilities in VLMs — arXiv:2610.09703 — arXiv (en), https://arxiv.org/abs/2610.09703 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_38583a7bf73e (Metraipolis archive dc_38583a7bf73e, sha256 60feb462313238c45863fd513b7bf2df50d7a89329fc381a29f002c874086052)
BibTeX
@misc{dc38583a7bf73e,
title = {Understanding and Mitigating Token-Pruning-Induced Vulnerabilities in VLMs — arXiv:2610.09703},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.09703},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_38583a7bf73e; sha256 60feb462313238c45863fd513b7bf2df50d7a89329fc381a29f002c874086052},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.09703. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Token-Pruning accelerates Vision-Language Models by removing redundant visual tokens, yet its safety implications remain underexplored. In this work, we present the first comprehensive safety evaluation of Token-Pruning mechanisms and find that: most pruning strategies significantly degrade safety as pruning ratios increase, whereas Query-based Compression shows the opposite, with extreme pruning (up to 99.8%), unexpectedly improves model safety. This sharp contrast prompts a key question: How do different Token-Pruning strategies reshape model safety behavior, and is it possible to enhance safety without sacrificing acceleration? To answer this, we identify an unrecognized mechanism, termed Pruning-Induced Malicious Amplification, where removal of background tokens triggers a side effect: forcing the model's attention to collapse onto a few retained malicious anchors within the foreground, inadvertently amplifying their toxic semantics under jailbreak. To address that, we propose an inference-time and plug-and-play Safety-Aware Pruning (SAP) mechanism that counteracts such dominance via three steps: (1) identifying malicious anchors, (2) restoring pruned benign tokens, and (3) reallocating excessive attention from malicious anchors to benign tokens. Extensive experiments across three safety and four utility benchmarks demonstrate that SAP mitigates pruning-induced vulnerabilities, i.e., reducing ASR by up to 62%, without compromising efficiency or utility. Tema degli esploratori: fiducia digitale e crittografia (elenco arXiv cs.CR, nuovi lavori del giorno). Categorie arXiv: Cryptography and Security (cs.CR) ; Machine Learning (cs.LG). Pagina del lavoro: https://arxiv.org/abs/2610.09703 — 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 2026-10-08T10:45:22.447Z dalla pagina https://arxiv.org/list/cs.CR/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).