Skip to yearly menu bar Skip to main content


Poster

Optimizing Watermarks for Large Language Models

Bram Wouters

Hall C 4-9 #2309
[ ] [ Paper PDF ]
[ Poster
Tue 23 Jul 4:30 a.m. PDT — 6 a.m. PDT

Abstract:

With the rise of large language models (LLMs) and concerns about potential misuse, watermarks for generative LLMs have recently attracted much attention. An important aspect of such watermarks is the trade-off between their identifiability and their impact on the quality of the generated text. This paper introduces a systematic approach to this trade-off in terms of a multi-objective optimization problem. For a large class of robust, efficient watermarks, the associated Pareto optimal solutions are identified and shown to outperform existing robust, efficient watermarks.

Chat is not available.