GumbelSoft: Diversified Language Model Watermarking via the GumbelMax-trick

Jiayi Fu, Xuandong Zhao, Ruihan Yang, Yuansen Zhang, Jiangjie Chen, Yanghua Xiao


Abstract
Large language models (LLMs) excellently generate human-like text, but also raise concerns about misuse in fake news and academic dishonesty. Decoding-based watermark, particularly the watermark based on the GumbelMax trick (GM watermark), is a standout solution for safeguarding machine-generated texts due to its notable detectability. However, GM watermark encounters a major challenge with generation diversity, always yielding identical outputs for the same prompt, negatively impacting generation diversity and user experience. To overcome this limitation, we introduce a new type of GM watermark, the Logits-Addition watermark, as well as three variants that aim to enhance diversity, particularly the GumbelSoft watermark (i.e., the softmax variant of the Logits-Addition watermark). When assessed for detectability in high diversity settings, our Gumbelsoft demonstrates superior performance, with its AUROC score exceeding those of the two alternative variants by a margin of 0.1 to 0.3 and outperforming other decoding-based watermarking methods by a minimum of 0.1.
Anthology ID:
2024.acl-long.315
Volume:
Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
Month:
August
Year:
2024
Address:
Bangkok, Thailand
Editors:
Lun-Wei Ku, Andre Martins, Vivek Srikumar
Venue:
ACL
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
5791–5808
Language:
URL:
https://aclanthology.org/2024.acl-long.315
DOI:
10.18653/v1/2024.acl-long.315
Bibkey:
Cite (ACL):
Jiayi Fu, Xuandong Zhao, Ruihan Yang, Yuansen Zhang, Jiangjie Chen, and Yanghua Xiao. 2024. GumbelSoft: Diversified Language Model Watermarking via the GumbelMax-trick. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 5791–5808, Bangkok, Thailand. Association for Computational Linguistics.
Cite (Informal):
GumbelSoft: Diversified Language Model Watermarking via the GumbelMax-trick (Fu et al., ACL 2024)
Copy Citation:
PDF:
https://aclanthology.org/2024.acl-long.315.pdf