@inproceedings{fu-etal-2024-gumbelsoft,
title = "{G}umbel{S}oft: Diversified Language Model Watermarking via the {G}umbel{M}ax-trick",
author = "Fu, Jiayi and
Zhao, Xuandong and
Yang, Ruihan and
Zhang, Yuansen and
Chen, Jiangjie and
Xiao, Yanghua",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = aug,
year = "2024",
address = "Bangkok, Thailand",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.acl-long.315",
doi = "10.18653/v1/2024.acl-long.315",
pages = "5791--5808",
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.",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="fu-etal-2024-gumbelsoft">
<titleInfo>
<title>GumbelSoft: Diversified Language Model Watermarking via the GumbelMax-trick</title>
</titleInfo>
<name type="personal">
<namePart type="given">Jiayi</namePart>
<namePart type="family">Fu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xuandong</namePart>
<namePart type="family">Zhao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ruihan</namePart>
<namePart type="family">Yang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yuansen</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiangjie</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yanghua</namePart>
<namePart type="family">Xiao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2024-08</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Lun-Wei</namePart>
<namePart type="family">Ku</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Andre</namePart>
<namePart type="family">Martins</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Vivek</namePart>
<namePart type="family">Srikumar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Bangkok, Thailand</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<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.</abstract>
<identifier type="citekey">fu-etal-2024-gumbelsoft</identifier>
<identifier type="doi">10.18653/v1/2024.acl-long.315</identifier>
<location>
<url>https://aclanthology.org/2024.acl-long.315</url>
</location>
<part>
<date>2024-08</date>
<extent unit="page">
<start>5791</start>
<end>5808</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T GumbelSoft: Diversified Language Model Watermarking via the GumbelMax-trick
%A Fu, Jiayi
%A Zhao, Xuandong
%A Yang, Ruihan
%A Zhang, Yuansen
%A Chen, Jiangjie
%A Xiao, Yanghua
%Y Ku, Lun-Wei
%Y Martins, Andre
%Y Srikumar, Vivek
%S Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
%D 2024
%8 August
%I Association for Computational Linguistics
%C Bangkok, Thailand
%F fu-etal-2024-gumbelsoft
%X 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.
%R 10.18653/v1/2024.acl-long.315
%U https://aclanthology.org/2024.acl-long.315
%U https://doi.org/10.18653/v1/2024.acl-long.315
%P 5791-5808
Markdown (Informal)
[GumbelSoft: Diversified Language Model Watermarking via the GumbelMax-trick](https://aclanthology.org/2024.acl-long.315) (Fu et al., ACL 2024)
ACL