Yepeng Liu

In-Context Watermarks for Large Language Models

Dataset Protection via Watermarked Canaries in Retrieval-Augmented LLMs

Theoretically Grounded Framework for LLM Watermarking: A Distribution-Adaptive Approach

Defending LLM Watermarking Against Spoofing Attacks with Contrastive Representation Learning

Image Watermarks are Removable using Controllable Regeneration from Clean Noise

Universally Optimal Watermarking Schemes for LLMs: from Theory to Practice