Optimizing Watermarks for Large Language Models

Bram Wouters
Proceedings of the 41st International Conference on Machine Learning, PMLR 235:53251-53269, 2024.

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.

Cite this Paper

BibTeX
@InProceedings{pmlr-v235-wouters24a, title = {Optimizing Watermarks for Large Language Models}, author = {Wouters, Bram}, booktitle = {Proceedings of the 41st International Conference on Machine Learning}, pages = {53251--53269}, year = {2024}, editor = {Salakhutdinov, Ruslan and Kolter, Zico and Heller, Katherine and Weller, Adrian and Oliver, Nuria and Scarlett, Jonathan and Berkenkamp, Felix}, volume = {235}, series = {Proceedings of Machine Learning Research}, month = {21--27 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v235/main/assets/wouters24a/wouters24a.pdf}, url = {https://proceedings.mlr.press/v235/wouters24a.html}, 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.} }
Endnote
%0 Conference Paper %T Optimizing Watermarks for Large Language Models %A Bram Wouters %B Proceedings of the 41st International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2024 %E Ruslan Salakhutdinov %E Zico Kolter %E Katherine Heller %E Adrian Weller %E Nuria Oliver %E Jonathan Scarlett %E Felix Berkenkamp %F pmlr-v235-wouters24a %I PMLR %P 53251--53269 %U https://proceedings.mlr.press/v235/wouters24a.html %V 235 %X 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.
APA
Wouters, B.. (2024). Optimizing Watermarks for Large Language Models. Proceedings of the 41st International Conference on Machine Learning, in Proceedings of Machine Learning Research 235:53251-53269 Available from https://proceedings.mlr.press/v235/wouters24a.html.

Related Material