Weiming Zhang

The Safeguard Worked. Is the LLM System Safer?

PURA: Provably Unbiased and Robust Multi-Bit Watermarking for AI-Generated Text Attribution

Breadcrumbing Search Agents

Safeguards Based on Copyable Context Cannot Provide Reliable Safety for LLMs

FARI: Robust One-Step Inversion for Watermarking in Diffusion Models

Between Safe Boundaries: Exploiting Temporal Consistency for Jailbreaking Text-To-Video Generation Models

Provably Secure Agent Guardrail

Into the Gray Zone: Domain Contexts Can Blur LLM Safety Boundaries

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

State-Dependent Safety Failures in Multi-Turn Language Model Interaction

Character as a Latent Variable in Large Language Models: A Mechanistic Account of Emergent Misalignment and Conditional Safety Failures

WMVLM: Evaluating Diffusion Model Image Watermarking via Vision-Language Models

SemBind: Binding Diffusion Watermarks to Semantics Against Black-Box Forgery Attacks

STEAD: Robust Provably Secure Linguistic Steganography with Diffusion Language Model