The paper introduces AttnPrint, a white‑box fingerprinting method that extracts low‑frequency components of cross‑modal attention distributions to identify multimodal large language models (MLLMs). It also presents DistillTrace, a black‑box auditing tool that uses hypothesis testing of MLLM outputs to detect potential model infringement. Experiments on 154 model instances across 19 architectures show that AttnPrint effectively detects derivative models and remains robust to downstream modifications, while DistillTrace reveals distillation relationships under various techniques.
arXiv:2608. 09928v1 Announce Type: cross Abstract: Multimodal Large Language Models (MLLMs) exhibit strong visual understanding, yet the internal features that cause these behaviors remain difficult to identify, audit, or control.
Vis-Poison is a novel attack that poisons multimodal retrieval-augmented generation systems by inserting attacker-controlled images as visual evidence, without altering any textual metadata. The attack uses an automated multi-agent approach to create visually plausible poisoned images and has been tested on two multimodal RAG pipelines, four embedding models, and six generation models. In black-box settings, Vis-Poison achieves an end-to-end success rate between 40.16% and 65.40% against 30,000-entry knowledge bases, and remains effective against various multimodal large language models with an average success rate above 60%.
The paper introduces CAIT, a benchmark of 400 synthetic scenes featuring counter‑intuitive actions that challenge multimodal large language models (MLLMs). Human participants and proprietary models like Claude and Gemini perform well, but standard open‑source instruction‑tuned MLLMs fail, largely due to a strong language prior that overrides contradictory visual evidence. The study shows that Chain‑of‑Thought reasoning can help but introduces new issues, while targeted fine‑tuning and structured prompting can reduce reliance on language priors and improve visual grounding.
arXiv:2511.18921v2 Announce Type: replace Abstract: Backdoor attacks undermine the reliability and trustworthiness of machine learning systems by injecting hidden behaviors that can be maliciously ac...
arXiv:2604. 05809v2 Announce Type: replace-cross Abstract: This paper presents Text-Guided Backdoor (TGB), an adjustable backdoor attack against multimodal pretrained models that uses natural-word triggers, namely words that can naturally occur in ordinary textual inputs.
ReFrame is a training‑free framework that enhances safety alignment for multimodal large language models at test time. It uses two lightweight agents: one generates risk and utility evidence, and the other rewrites prompts and routes images to create a safe proxy before invoking the deployed MLLM. Experiments show that ReFrame improves jailbreak defense, safety awareness, and reduces over‑sensitivity while maintaining multimodal utility.
The paper introduces a new adversarial attack on multimodal large language models (MLLMs) that targets the models’ intrinsic focus in the frequency domain. By exploiting the models’ sensitivity to phase information, the authors design a phase‑aware perturbation strategy that restricts changes to structure‑relevant phase regions, making attacks both effective and imperceptible. An auxiliary adversarial prompt module further misaligns multimodal attention toward targeted structural patterns, and experiments on several MLLM models and datasets confirm the method’s superior performance over existing attacks.
The paper investigates how different multimodal design choices affect the performance of misinformation detection systems. Using over 3,375 experiments across three benchmark datasets and various pre‑trained vision and language models, the authors systematically compare design options and conduct robustness analyses. The study offers practical guidance on which choices improve detection, when they may fail silently, and which pipeline components most influence model behavior, addressing four key research questions.
The paper investigates cross‑modal safety drift in multimodal large language models, where a harmless text query paired with a visual image can trigger harmful responses. Empirical analysis identifies unsafe response patterns and shows that visual cues receive limited attention, weakening refusal mechanisms. The authors introduce Safety‑Awareness Representation Transfer (SRT), a lightweight method that transfers safety signals from text processing to mitigate cross‑modal drift while maintaining model utility.
arXiv:2610.00341v1 Announce Type: cross Abstract: As Large Multimodal Models (LMMs) transition toward natively unified architectures, evaluating their safety in synergistic harmful image-text generat...
arXiv:2608. 06865v1 Announce Type: cross Abstract: The malicious use of generative artificial intelligence to create highly realistic deepfake videos raises serious ethical concerns and poses substantial challenges to AI safety.
Vision-Language Large Models (VLLMs) trained on massive crawled corpora raise pressing copyright and data-provenance concerns. These concerns are particularly acute in healthcare, where patient medical images paired with clinical reports demand rigorous privacy safeguards.