English
Related papers

Related papers: When Think-with-Image Meets Safety: What Determine…

200 papers

Building on the success of text-based reasoning models like DeepSeek-R1, extending these capabilities to multimodal reasoning holds great promise. While recent works have attempted to adapt DeepSeek-R1-style reinforcement learning (RL)…

Computer Vision and Pattern Recognition · Computer Science 2025-06-10 Jie Yang , Feipeng Ma , Zitian Wang , Dacheng Yin , Kang Rong , Fengyun Rao , Ruimao Zhang

With the growing deployment of Vision-Language Models (VLMs) in real-world applications, previously overlooked safety risks are becoming increasingly evident. In particular, seemingly innocuous multimodal inputs can combine to reveal…

Computer Vision and Pattern Recognition · Computer Science 2025-10-07 Youngjin Na , Sangheon Jeong , Youngwan Lee , Jian Lee , Dawoon Jeong , Youngman Kim

The vulnerability of Vision Large Language Models (VLLMs) to jailbreak attacks appears as no surprise. However, recent defense mechanisms against these attacks have reached near-saturation performance on benchmark evaluations, often with…

Cryptography and Security · Computer Science 2025-03-07 Yangyang Guo , Fangkai Jiao , Liqiang Nie , Mohan Kankanhalli

Optimization-based jailbreaks typically adopt the Toxic-Continuation setting in large vision-language models (LVLMs), following the standard next-token prediction objective. In this setting, an adversarial image is optimized to make the…

Computer Vision and Pattern Recognition · Computer Science 2025-05-29 Hee-Seon Kim , Minbeom Kim , Wonjun Lee , Kihyun Kim , Changick Kim

Vision-Language-Action (VLA) models with integrated reasoning have been proposed for end-to-end autonomous driving, assuming a tight coupling between reasoning and trajectory generation. However, the robustness of such systems under…

Cryptography and Security · Computer Science 2026-05-29 Mohammadreza Teymoorianfard , Jean-Philippe Monteuuis , Jonathan Petit , Amir Houmansadr

Research endeavors have been made in learning robust Multimodal Large Language Models (MLLMs) against jailbreak attacks. However, existing methods for improving MLLMs' robustness still face critical challenges: \ding{172} how to efficiently…

Computer Vision and Pattern Recognition · Computer Science 2026-01-07 Liming Lu , Xiang Gu , Shuchao Pang , Siyuan Liang , Haotian Zhu , Xiyu Zeng , Xu Zheng , Yongbin Zhou

Jailbreak attacks pose a serious threat to Large Language Models (LLMs) by bypassing their safety mechanisms. A truly advanced jailbreak is defined not only by its effectiveness but, more critically, by its stealthiness. However, existing…

Cryptography and Security · Computer Science 2026-03-13 Jianing Geng , Biao Yi , Zekun Fei , Ruiqi He , Lihai Nie , Tong Li , Zheli Liu

With the rapid advancements in Multimodal Large Language Models (MLLMs), securing these models against malicious inputs while aligning them with human values has emerged as a critical challenge. In this paper, we investigate an important…

Cryptography and Security · Computer Science 2024-11-26 Weidi Luo , Siyuan Ma , Xiaogeng Liu , Xiaoyu Guo , Chaowei Xiao

Large language models (LLMs) have become increasingly integrated with various applications. To ensure that LLMs do not generate unsafe responses, they are aligned with safeguards that specify what content is restricted. However, such…

Computation and Language · Computer Science 2024-05-08 Hongyu Cai , Arjun Arunasalam , Leo Y. Lin , Antonio Bianchi , Z. Berkay Celik

Large Vision-Language Models (LVLMs) are vulnerable to a growing array of multimodal jailbreak attacks, necessitating defenses that are both generalizable to novel threats and efficient for practical deployment. Many current strategies fall…

Cryptography and Security · Computer Science 2026-04-21 Peichun Hua , Hao Li , Shanghao Shi , Zhiyuan Yu , Ning Zhang

Existing attacks against multimodal language models (MLLMs) primarily communicate instructions through text accompanied by adversarial images. In contrast, we exploit the capabilities of MLLMs to interpret non-textual instructions,…

Cryptography and Security · Computer Science 2025-06-03 Jiahui Geng , Thy Thy Tran , Preslav Nakov , Iryna Gurevych

The safety alignment of large language models (LLMs) remains vulnerable, as their initial behavior can be easily jailbroken by even relatively simple attacks. Since infilling a fixed template between the input instruction and initial model…

Computation and Language · Computer Science 2025-06-05 Chak Tou Leong , Qingyu Yin , Jian Wang , Wenjie Li

Many-shot jailbreaking (MSJ) is an adversarial technique that exploits the long context windows of modern LLMs to circumvent model safety training by including in the prompt many examples of a "fake" assistant responding inappropriately…

Machine Learning · Computer Science 2026-03-26 Christopher M. Ackerman , Nina Panickssery

Reinforcement Learning with Verifiable Rewards (RLVR) has significantly advanced the reasoning capabilities of Multimodal Large Language Models (MLLMs), yet how visual evidence is integrated during reasoning remains poorly understood. We…

Artificial Intelligence · Computer Science 2026-02-13 Zhengbo Jiao , Shaobo Wang , Zifan Zhang , Wei Wang , Bing Zhao , Hu Wei , Linfeng Zhang

Large Language Models have found success in a variety of applications. However, their safety remains a concern due to the existence of various jailbreaking methods. Despite significant efforts, alignment and safety fine-tuning only provide…

Computation and Language · Computer Science 2025-12-16 Darpan Aswal , Céline Hudelot

Safety is critical to the usage of large language models (LLMs). Multiple techniques such as data filtering and supervised fine-tuning have been developed to strengthen LLM safety. However, currently known techniques presume that corpora…

Computation and Language · Computer Science 2024-06-10 Fengqing Jiang , Zhangchen Xu , Luyao Niu , Zhen Xiang , Bhaskar Ramasubramanian , Bo Li , Radha Poovendran

Jailbreak attacks, where harmful prompts bypass generative models' built-in safety, raise serious concerns about model vulnerability. While many defense methods have been proposed, the trade-offs between safety and helpfulness, and their…

Cryptography and Security · Computer Science 2025-02-21 Zhuohang Long , Siyuan Wang , Shujun Liu , Yuhang Lai , Xuanjing Huang , Zhongyu Wei

The visual modality of vision-language models (VLMs) is an underexplored attack surface for bypassing safety alignment. We introduce four jailbreak attacks exploiting the vision component: (1) encoding harmful instructions as visual symbol…

Computer Vision and Pattern Recognition · Computer Science 2026-05-04 Aharon Azulay , Jan Dubiński , Zhuoyun Li , Atharv Mittal , Yossi Gandelsman

Despite extensive safety-tuning, large language models (LLMs) remain vulnerable to jailbreak attacks via adversarially crafted instructions, reflecting a persistent trade-off between safety and task performance. In this work, we propose…

Cryptography and Security · Computer Science 2025-08-26 Wei Jie Yeo , Ranjan Satapathy , Erik Cambria

This study reveals a critical safety blind spot in modern LLMs: learning-style queries, which closely resemble ordinary educational questions, can reliably elicit harmful responses. The learning-style queries are constructed by a novel…

Cryptography and Security · Computer Science 2026-02-25 Xuan Luo , Yue Wang , Zefeng He , Geng Tu , Jing Li , Ruifeng Xu