Publications

Reading Between the Pixels: An Inscriptive Jailbreak Attack on Text-to-Image Models

SafeGen: Goal-Conditioned Video Diffusion of Safety-Critical Scenarios for VLM-Based Autonomous Driving

DeltaUI: Framework-Normalized UI State Transition Modeling for Multi-Task Front-End Engineering

SafeHarbor: Hierarchical Memory-Augmented Guardrail for LLM Agent Safety

DMN: A Compositional Framework for Jailbreaking Multimodal LLMs with Multi-Image Inputs

AGENTSAFE: Benchmarking the Safety of Embodied Agents on Hazardous Instructions

Delegated Misalignment: How Multi-Agent Structures Amplify LLM Safety Risks

Causal Abstention for Cost-Aware Language-Agent Collaboration

SecureWebArena: A Holistic Security Evaluation Benchmark for LVLM-based Web Agents

Uncovering Strategic Egoism Behaviors in Large Language Models

RoboSafe: Safeguarding Embodied Agents via Executable Safety Logic

Mask-GCG: Are All Tokens in Adversarial Suffixes Necessary for Jailbreak Attacks?

Probabilistic Modeling of Jailbreak on Multimodal LLMs: From Quantification to Application

Robust Rumor Detection Against Noise

DIVER: Dynamic Iterative Visual Evidence Reasoning for Multimodal Fake News Detection

Reasoning-Oriented Programming: Chaining Semantic Gadgets to Jailbreak Large Vision Language Models

Uncovering Security Threats and Architecting Defenses in Autonomous Agents: A Case Study of OpenClaw

Evolving Deception: When Agents Evolve, Deception Wins

AgentVisor: Defending LLM Agents Against Prompt Injection via Semantic Virtualization

GuardAD: Safeguarding Autonomous Driving MLLMs via Markovian Safety Logic

TrajShield: Trajectory-Level Safety Mediation for Defending Text-to-Video Models Against Jailbreak Attacks

Detoxifying Large Language Models via Autoregressive Reward Guided Representation Editing

Manipulating Multimodal Agents via Cross-Modal Prompt Injection

Reasoning-Augmented Conversation for Multi-Turn Jailbreak Attacks on Large Language Models

Jailbreak Vision Language Models via Bi-Modal Adversarial Prompt

CogMorph: Cognitive Morphing Attacks for Text-to-Image Models

PRJ: Perception–Retrieval–Judgement for Generated Images

SafeBench: A Safety Evaluation Framework for Multimodal Large Language Models

Towards Understanding the Safety Boundaries of DeepSeek Models: Evaluation and Findings

Pushing the Limits of Safety: A Technical Report on the ATLAS Challenge 2025

Prism: Programmatic Reasoning with Image Sequence Manipulation for LVLM Jailbreaking

VEIL: Jailbreaking Text-to-Video Models via Visual Exploitation from Implicit Language

Sequential Comics for Jailbreaking Multimodal Large Language Models via Structured Visual Storytelling

Disentangling Fact from Sentiment: A Dynamic Conflict-Consensus Framework for Multimodal Fake News Detection

Utilizing Jailbreak Probability to Attack and Safeguard Multimodal LLMs

Unveiling the Safety of GPT-4o: An Empirical Study using Jailbreak Attacks

CS-Eval: A Comprehensive Large Language Model Benchmark for Cybersecurity

NBA: Defensive Distillation for Backdoor Removal via Neural Behavior Alignment

DLP: Towards Active Defense Against Backdoor Attacks with Decoupled Learning Process

DeeSCVHunter: A Deep Learning-Based Framework for Smart Contract Vulnerability Detection