10 papers
Every Picture Tells a Dangerous Story: Memory-Augmented Multi-Agent Jailbreak Attacks on VLMs
Jianhao Chen, Haoyang Chen, Shiqin Wang +4
Vision-Language Models (VLMs) expand the attack surface of safety-aligned systems by coupling visual perception with text generation. Existing multimodal jailbreak attacks primaril…
Controlling Multimodal Conversational Agents with Coverage-Enhanced Latent Actions
Yongqi Li, Hao Lang, Tieyun Qian +1
Vision-language models are increasingly employed as multimodal conversational agents (MCAs) for diverse conversational tasks. Recently, reinforcement learning (RL) has been widely…
Understanding Generalization in Role-Playing Models via Information Theory
Yongqi Li, Hao Lang, Fei Huang +2
Role-playing models (RPMs) are widely used in real-world applications but underperform when deployed in the wild. This degradation can be attributed to distribution shifts, includi…
ContiGuard: A Framework for Continual Toxicity Detection Against Evolving Evasive Perturbations
Hankun Kang, Xin Miao, Jianhao Chen +5
Toxicity detection mitigates the dissemination of toxic content (e.g., hateful comments, posts, and messages within online social actions) to safeguard a healthy online social envi…
RAJ-PGA: Reasoning-Activated Jailbreak and Principle-Guided Alignment Framework for Large Reasoning Models
Jianhao Chen, Mayi Xu, Haoyang Chen +6
Large Reasoning Models (LRMs) face a distinct safety vulnerability: their internal reasoning chains may generate harmful content even when the final output appears benign. To addre…
CAVGAN: Unifying Jailbreak and Defense of LLMs via Generative Adversarial Attacks on their Internal Representations
Xiaohu Li, Yunfeng Ning, Zepeng Bao +3
Security alignment enables the Large Language Model (LLM) to gain the protection against malicious queries, but various jailbreak attack methods reveal the vulnerability of this se…