Shaopeng Fu

Understanding and Improving Continuous LLM Adversarial Training via In-context Learning Theory

Benign Overfitting in Adversarial Training for Vision Transformers

Understanding Private Learning From Feature Perspective

Pre-trained Encoder Inference: Revealing Upstream Encoders In Downstream Machine Learning Services

Concept-Based Dictionary Learning for Inference-Time Safety in Vision–Language–Action Models

Backdooring CLIP through Concept Confusion

Short-length Adversarial Training Helps LLMs Defend Long-length Jailbreak Attacks: Theoretical and Empirical Evidence

Representation Confusion: Towards Representation Backdoor on CLIP via Concept Activation