Wenbo Zhou

VLMShield: Efficient and Robust Defense of Vision-Language Models against Malicious Prompts

Revoking Amnesia: RL-based Trajectory Optimization to Resurrect Erased Concepts in Diffusion Models

Patch-Wise and Keyword-Aware: Efficient Multi-Condition Control of Diffusion Transformers via Position-Aligned and Keyword-Scoped Attention

Talk2Me: High-Fidelity and Controllable Audio-Driven Avatars with Gaussian Splatting

MoLD: Fine-Grained Multimodal Risk Assessment via Dynamic Analysis Weights

T2V-OptJail: Discrete Prompt Optimization for Text-to-Video Jailbreak Attacks

EraseAnything: Enabling Concept Erasure in Rectified Flow Transformers