14 papers
Invisible Ink, Visible Lies: How Production Watermarking Causes LLMs to Hallucinate
Haocheng Ye, Aoting Hu, Xinwei Zhang +3
Text watermarking helps identify AI-generated content, but its effect on factual reliability remains underexplored. In this paper, we study watermarking hallucination: factual erro…
SteerProbe: Learning to Bypass Safety Steering in Vision-Language Models
Xinwei Zhang, Aoting Hu, Hangcheng Liu +3
Activation steering offers an inference-time defense for vision--language models (VLMs) by modifying intermediate representations without updating backbone parameters. However, pro…
AI Alignment through a Game-theoretic Lens: A Survey
Yanan Cai, Zhongrui Zhao, Zhigang Lu +6
As large language models and increasingly capable AI agents are deployed in high-risk settings, aligning them with complex human values has become a central challenge. Existing ali…
Toward Trustworthy AI: Multi-Target Adversarial Attacks and Robust Defenses for Continuous Data Summarization
Yuefang Lian, Longkun Guo, Zhongrui Zhao +5
Trustworthy AI requires reliable data-processing pipelines, not only robust downstream predictive models. As an upstream component, data summarization determines which information…
If you're waiting for a sign... that might not be it! Mitigating Trust Boundary Confusion from Visual Injections on Vision-Language Agentic Systems
Jiamin Chang, Minhui Xue, Ruoxi Sun +3
Recent advances in embodied Vision-Language Agentic Systems (VLAS), powered by large vision-language models (LVLMs), enable AI systems to perceive and reason over real-world scenes…
AgentRAE: Remote Action Execution through Notification-based Visual Backdoors against Screenshots-based Mobile GUI Agents
Yutao Luo, Haotian Zhu, Shuchao Pang +4
The rapid adoption of mobile graphical user interface (GUI) agents, which autonomously control applications and operating systems (OS), exposes new system-level attack surfaces. Ex…
DREAM: Dynamic Red-teaming across Environments for AI Models
Liming Lu, Xiang Gu, Junyu Huang +5
Large Language Models (LLMs) are increasingly used in agentic systems, where their interactions with diverse tools and environments create complex, multi-stage safety challenges. H…
Multimodal Robust Prompt Distillation for 3D Point Cloud Models
Xiang Gu, Liming Lu, Xu Zheng +3
Adversarial attacks pose a significant threat to learning-based 3D point cloud models, critically undermining their reliability in security-sensitive applications. Existing defense…
RoboView-Bias: Benchmarking Visual Bias in Embodied Agents for Robotic Manipulation
Enguang Liu, Siyuan Liang, Liming Lu +4
The safety and reliability of embodied agents rely on accurate and unbiased visual perception. However, existing benchmarks mainly emphasize generalization and robustness under per…
FERD: Fairness-Enhanced Data-Free Robustness Distillation
Zhengxiao Li, Liming Lu, Xu Zheng +4
Data-Free Robustness Distillation (DFRD) aims to transfer the robustness from the teacher to the student without accessing the training data. While existing methods focus on overal…
CIARD: Cyclic Iterative Adversarial Robustness Distillation
Liming Lu, Shuchao Pang, Xu Zheng +4
Adversarial robustness distillation (ARD) aims to transfer both performance and robustness from teacher model to lightweight student model, enabling resilient performance on resour…
Towards a 3D Transfer-based Black-box Attack via Critical Feature Guidance
Shuchao Pang, Zhenghan Chen, Shen Zhang +4
Deep neural networks for 3D point clouds have been demonstrated to be vulnerable to adversarial examples. Previous 3D adversarial attack methods often exploit certain information a…
EAT: Multimodal Jailbreak Defense via Dynamic Joint Optimization for Multimodal Large Language Models
Liming Lu, Xiang Gu, Shuchao Pang +5
Research endeavors have been made in learning robust Multimodal Large Language Models (MLLMs) against jailbreak attacks. However, existing methods for improving MLLMs' robustness s…
Reconstruction of Differentially Private Text Sanitization via Large Language Models
Shuchao Pang, Zhigang Lu, Haichen Wang +3
Differential privacy (DP) is the de facto privacy standard against privacy leakage attacks, including many recently discovered ones against large language models (LLMs). However, w…