Metraipolis › archive › dc_7b5fba4e7888
Counterfactual Neighborhood Risk Fields: Evidence-Grounded Selective Safety for Vision-Language Models — arXiv:2610.20679
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.20679
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-16T08:59:25.241Z
- SHA-256 of the archived text
bc27bfc6015e9c7faf1358a9bc5cb0adbb4108bbf4ef4313a40d8ef07be4fe91- Stable URL of this page
https://metraipolis.com/documento/dc_7b5fba4e7888
How to cite
Counterfactual Neighborhood Risk Fields: Evidence-Grounded Selective Safety for Vision-Language Models — arXiv:2610.20679 — arXiv (en), https://arxiv.org/abs/2610.20679 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_7b5fba4e7888 (Metraipolis archive dc_7b5fba4e7888, sha256 bc27bfc6015e9c7faf1358a9bc5cb0adbb4108bbf4ef4313a40d8ef07be4fe91)
BibTeX
@misc{dc7b5fba4e7888,
title = {Counterfactual Neighborhood Risk Fields: Evidence-Grounded Selective Safety for Vision-Language Models — arXiv:2610.20679},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.20679},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_7b5fba4e7888; sha256 bc27bfc6015e9c7faf1358a9bc5cb0adbb4108bbf4ef4313a40d8ef07be4fe91},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.20679. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Representation-based safety defenses for vision-language models (VLMs) often learn to separate harmful from benign inputs, but separation alone does not establish what a risk score means or when it should be trusted. We introduce Counterfactual Neighborhood Risk Fields (CNRF), an evidence-grounded framework for selective VLM safety. CNRF constructs context-matched benign-harmful counterfactual pairs, uses their representation differences as capability-related risk directions, and estimates neighborhood support separately from risk. This separation prevents unfamiliarity from being conflated with harmfulness and enables selective intervention rather than blanket rejection. A victim-adaptive interface accommodates different safety observations across VLM architectures and routes selected requests for same-victim response review, which either releases acceptable candidates or produces constrained alternatives with reduced harmful capability. Across three VLMs and five multimodal jailbreak families, CNRF achieves 86.06-89.30% macro defense success rate while retaining substantial benign-task utility. Controlled experiments further show that query-matched references and preserved context-direction correspondence materially improve risk ranking in the low-false-positive regime. CNRF thus reframes representation-based VLM safety from learning harmfulness boundaries to constructing applicable evidence for selective risk decisions. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno). Categorie arXiv: Artificial Intelligence (cs.AI) ; Cryptography and Security (cs.CR); Multimedia (cs.MM). Pagina del lavoro: https://arxiv.org/abs/2610.20679 — 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 2026-10-16T08:44:17.646Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).