1 paper · 1 filter
Georgios Milis, Yubin Qin, Yihan Wu +1
As policy catches up with the capabilities of generative AI, watermarking is central to content provenance efforts. Inference-time watermarks for autoregressive models are unfit fo…