Metraipolis › archive › dc_87c93ca0b6ce
SafeRI: Recognition and Intervention for Token-Level Safety Intervention in Large Vision Language Models — arXiv:2609.03544
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.03544
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-17T02:56:17.790Z
- SHA-256 of the archived text
4bb1ca84b8f7c8ac980b72409d19897694afbfa65e937e7b381fc52cb36d7c51- Stable URL of this page
https://metraipolis.com/documento/dc_87c93ca0b6ce
How to cite
SafeRI: Recognition and Intervention for Token-Level Safety Intervention in Large Vision Language Models — arXiv:2609.03544 — arXiv (en), https://arxiv.org/abs/2609.03544 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_87c93ca0b6ce (Metraipolis archive dc_87c93ca0b6ce, sha256 4bb1ca84b8f7c8ac980b72409d19897694afbfa65e937e7b381fc52cb36d7c51)
BibTeX
@misc{dc87c93ca0b6ce,
title = {SafeRI: Recognition and Intervention for Token-Level Safety Intervention in Large Vision Language Models — arXiv:2609.03544},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.03544},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_87c93ca0b6ce; sha256 4bb1ca84b8f7c8ac980b72409d19897694afbfa65e937e7b381fc52cb36d7c51},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.03544. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Existing safety alignment methods for vision-language models usually modify the model behavior globally: once the safety parameters are trained or loaded, they participate in both unsafe and already-safe generations. This always-on intervention can unnecessarily perturb the model's original reasoning path and degrade general multimodal capabilities. We argue that safety alignment should be an on-demand intervention rather than a permanent modification to every decoding trajectory. To this end, we propose a streaming recognition and gated LoRA framework for intrinsic VLM safety. During autoregressive generation, a lightweight recognizer estimates whether the current pre-token generation state is safe or unsafe. Its output updates the LoRA gate for the following decoding step; otherwise, generation follows the frozen-backbone policy. The LoRA module is trained from unsafe prefixes, transition statements, and safe continuations, so that it learns to redirect unsafe generations back to safe responses after activation. Experiments across multiple safety and general-purpose benchmarks demonstrate the effectiveness of our method in post-alignment settings. Tema degli esploratori: statistica e scienza dei dati (retro-catalogo arXiv: annunci del 2026-09-04 in cs.CV). Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV). Pagina del lavoro: https://arxiv.org/abs/2609.03544 — 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 2026-10-17T02:27:18.968Z dalla pagina https://arxiv.org/catchup/cs.CV/2026-09-04?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).