Metraipolis › archive › dc_8a2969a582c6
When Reasoning Is Not Enough: Proactive Inquiry for Cross-Modal Safety in Multimodal Large Language Models — arXiv:2610.22068
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.22068
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-16T09:05:34.634Z
- SHA-256 of the archived text
5c42070fb97c4a18bff2dbedd3094ec03d662d7a4af9566bc51a5347b453d9a3- Stable URL of this page
https://metraipolis.com/documento/dc_8a2969a582c6
How to cite
When Reasoning Is Not Enough: Proactive Inquiry for Cross-Modal Safety in Multimodal Large Language Models — arXiv:2610.22068 — arXiv (en), https://arxiv.org/abs/2610.22068 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_8a2969a582c6 (Metraipolis archive dc_8a2969a582c6, sha256 5c42070fb97c4a18bff2dbedd3094ec03d662d7a4af9566bc51a5347b453d9a3)
BibTeX
@misc{dc8a2969a582c6,
title = {When Reasoning Is Not Enough: Proactive Inquiry for Cross-Modal Safety in Multimodal Large Language Models — arXiv:2610.22068},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.22068},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_8a2969a582c6; sha256 5c42070fb97c4a18bff2dbedd3094ec03d662d7a4af9566bc51a5347b453d9a3},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.22068. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Multimodal large language models (MLLMs) can act unsafely when individually benign images and texts are interpreted jointly. Beyond implicit reasoning risk, where reasoning drifts toward harm, we identify intent-level risk, where cross-modal ambiguity silently commits the model to a hazardous premise that path-level optimization alone cannot correct. We propose SAFE-PRO, which uses a Risk-Conditioned Uncertainty (RCU) signal, combining predictive entropy with a learned cross-modal risk score, to decide when to ask safety-grounded clarifying questions and where to apply path optimization, under a reward that prioritizes safety and penalizes over-refusal. We also build SSUI-Dialog, a multimodal multi-turn dataset with dual-intent supervision, and theoretically analyze gradient variance and unsafe-completion risk under RCU. Across seven benchmarks and two base MLLMs, SAFE-PRO consistently improves safety over SRPO, roughly halves its over-refusal, and surpasses GPT-4o and Claude-3.5-Sonnet in average single-turn safety. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno). Categorie arXiv: Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2610.22068 — 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 2026-10-16T08:44:17.809Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).