The paper investigates how Vision Language Models (VLMs) can be fooled by tiny, human‑imperceptible changes to images. It introduces a gradient‑based attack that targets only the vision encoder, reducing computational cost while still effectively disrupting both untargeted and targeted multimodal interpretations. Experiments on open‑source VLMs such as Qwen2.5‑VL, Granite‑Vision, FastVLM, and Phi‑3.5‑Vision demonstrate that these small perturbations can dramatically alter the models’ textual outputs.
The paper investigates how Vision Language Models (VLMs) can be fooled by small, human‑imperceptible changes to images. It introduces a gradient‑based attack that targets only the vision encoder, reducing computational cost while still effectively disrupting both untargeted and targeted multimodal alignment. Experiments on open‑source VLMs such as Qwen2.5‑VL, Granite‑Vision, FastVLM, and Phi‑3.5‑Vision demonstrate that these perturbations can significantly alter the models’ textual outputs.
By Ilan Zini, Boussad Addad, Katarzyna Kapusta
arXiv:2512. 21815v4 Announce Type: replace-cross Abstract: Vision-language models (VLMs) achieve remarkable performance but remain vulnerable to adversarial attacks.
By Mengqi He, Xinyu Tian, Xin Shen, Jinhong Ni, Shu Zou, Zhaoyuan Yang, Jing Zhang
arXiv:2606. 15308v1 Announce Type: new Abstract: While multimodal large language models (MLLMs) have shown strong visual reasoning abilities, serving a large model for every query is computationally expensive.
By Zhongye Liu, Yaopei Zeng, Yurui Chang, Lu Lin
arXiv:2606. 05958v1 Announce Type: new Abstract: Activation steering has become a popular way to control Large Language Model (LLM) behavior without fine-tuning.
By Abzal Aidakhmetov, Donato Crisostomi, Tommaso Mencattini, Adrian Robert Minut, Iacopo Masi, Emanuele Rodol\`a
arXiv:2606. 24388v1 Announce Type: new Abstract: We introduce a large-scale, open-source dataset of pre-generated adversarial attacks for vision-language models (VLMs).
By Simone Gallivanone, Hossein Khodadadi, Mauro Dore, Mauro Medda, Nicola Franco
The paper introduces TA-SPA, a black‑box jailbreak method for multimodal large language models that generates transferable perturbations in a text‑anchored semantic space. It combines Text‑Anchored Semantic Factorization (TASF) to separate cross‑modal semantic factors from modality‑specific residuals with Semantic‑Preserving Augmentation (SPA) to diversify harmful target anchors while maintaining semantic consistency. Experiments demonstrate strong attack effectiveness and transferability to commercial MLLMs, with competitive performance against representative defenses.
By Wenyun Li, Guiping Cao, Xiangyuan Lan, Zheng Zhang
arXiv:2609.25049v2 Announce Type: replace-cross
Abstract: Large language models (LLMs) aligned for safety often suffer from over-refusal, incorrectly rejecting benign yet safety-related instructions....
By Zixuan Wang, Bingjie Zhang, He Zhao, Dandan Guo
FSPGD introduces a feature-space black-box attack for semantic segmentation that targets intermediate representations rather than just output logits. The method uses a dual loss: an external loss to disrupt cross-model feature alignment and an internal loss to reduce consistency among same-class instances. Experiments on Pascal VOC 2012 and Cityscapes show that FSPGD outperforms existing logit-level and segmentation-specific attacks across CNN and Transformer backbones, and its adversarial examples improve robustness when used for training.
By Eun-Sol Park, MiSo Park, Yong-Goo Shin
The paper addresses the problem of over‑refusal in safety‑aligned large language models, where benign instructions are incorrectly rejected. It identifies that a small set of hypersensitive safety heads in transformer attention misfire on hard‑safe prompts, causing abnormal attention entanglement and high‑entropy routing conflicts that block necessary attention to target entities. To mitigate this, the authors propose Semantic Routing Calibration (SRC), a lightweight, training‑free inference framework that dynamically suppresses these hypersensitive heads and fuses logits from dual branches to restore trustworthy reasoning while preserving intrinsic safety performance.
By Zixuan Wang, Bingjie Zhang, He Zhao, Dandan Guo
Multimodal Large Language Models (MLLMs) have achieved remarkable progress in vision-language interaction, yet their safety alignment remains vulnerable to jailbreak attacks. A key challenge is that s...
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...
By Juncheng Li, Yige Li, Hanxun Huang, Yunhao Chen, Xin Wang, Yixu Wang, Xingjun Ma, Yu-Gang Jiang