Problem
Research questions and friction points this paper is trying to address.
Multimodal Large Language Models
Safety Risk
Reference-Dependent Failure Mode
Operation-Target Pair
Visual Reference
COMIC: Reference-Aware Safety Gating for Multimodal Large Language Models
🤖 AI Summary
本文针对多模态大语言模型在处理视觉定位输入时的安全风险，提出了一种名为COMIC的参考感知安全门控方法，有效提升了模型的安全性。
📝 Abstract
Multimodal large language models (MLLMs) are increasingly used to interact with screenshots, scanned documents, diagrams, and other visually grounded inputs. This shift introduces a new safety risk: in many multimodal jailbreaks, neither the prompt nor the image is harmful in isolation. Unsafe behavior emerges only when the model binds an apparently benign operation, such as summarizing, translating, or following, to a localized visual target. This reveals a structural weakness in current multimodal defenses, which largely moderate the prompt-image pair as a whole even though the true security-relevant unit is the grounded operation-target pair produced during dereference. In this work, we identify and analyze this reference-dependent failure mode and show that existing defenses degrade when harmful semantics are localized, activated only after grounding, and dependent on visual reference resolution. To address this problem, we propose COMIC (Context-Operation-Modality-Image-Classifier), a reference-aware pre-generation safety gate for MLLMs. COMIC first infers the requested operation and reference type, constructs candidate targets from OCR and open-vocabulary proposals, grounds plausible referents, and evaluates safety over explicit operation-target pairs. To handle ambiguity conservatively, COMIC combines max-risk aggregation with quality-aware routing before deciding whether to forward or block a request. We evaluate COMIC across multiple open-source MLLMs, localized and broader multimodal jailbreak benchmarks, and benign reference-sensitive settings. The results show that COMIC consistently improves robustness while preserving benign utility and practical efficiency. More broadly, our findings suggest that multimodal safety cannot be enforced reliably without modeling the requested operation, the visual target to which it applies, and the confidence of that grounding.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Reference-Aware
Safety Gating
Multimodal Large Language Models
Visual Reference Resolution
Max-Risk Aggregation
🔎 Similar Papers
No similar papers found.
