Jie Zhang

Black-box Optimization of LLM Outputs by Asking for Directions

The Jailbreak Tax: How Useful are Your Jailbreak Outputs?

Textual Unlearning Gives a False Sense of Unlearning

Does Training with Synthetic Data Truly Protect Privacy?

Prompt-Independent Safe Decoding to Restrain Unsafe Image Generation for Text-to-Image Models against White-Box Adversary

Blind Baselines Beat Membership Inference Attacks for Foundation Models