大模型
arxiv.org ·
多语言 LLM 水印真的是多语言的么？通过回译将鲁棒性扩展至 100 多种语言
该论文指出，现有多语言 LLM 水印方法虽声称跨语言鲁棒，但仅在高资源语言上评估，在中低资源语言遭遇翻译攻击时失效，原因在于分词器对相应语言缺少足够全词 token，导致语义聚类失败。作者提出 STEAM：用贝叶斯优化在 126 种候选语言中搜索能最大程度恢复水印强度的回译，兼容任意水印方法，跨分词器与语言鲁棒、非侵入且易扩展。报告平均提升 +0.23 AUC 与 +37%p TPR@1%。
AI推荐理由 arXiv 论文（v3，2026-09-11 修订），问题明确、方法具体：指出多语言 LLM 水印在翻译攻击下对中低资源语言不鲁棒，归因于语义聚类在分词器全词 token 不足时失效，并提出 STEAM——用贝叶斯优化在 126 种候选语言中搜索最佳回译以恢复水印强度，报告平均 +0.23 AUC、+37%p TPR@1%。新信息在于对现有方法局限的实证指认与可扩展检测方案；局限是摘要未给出数据集、基线细节与复现条件，增益为作者自报。
78
正文
Computer Science > Computation and Language
arXiv:2510.18019 (cs)
[Submitted on 20 Oct 2025 (v1), last revised 11 Sep 2026 (this version, v3)]
Title:Is Multilingual LLM Watermarking Truly Multilingual? Scaling Robustness to 100+ Languages via Back-Translation
Authors:Asim Mohamed, Martin Gubri
View a PDF of the paper titled Is Multilingual LLM Watermarking Truly Multilingual? Scaling Robustness to 100+ Languages via Back-Translation, by Asim Mohamed and Martin Gubri
View PDF
HTML (experimental)
Abstract:Multilingual watermarking aims to make large language model (LLM) outputs traceable across languages, yet current methods still fall short. Despite claims of cross-lingual robustness, they are evaluated only on high-resource languages. We show that existing multilingual watermarking methods are not truly multilingual: they fail to remain robust under translation attacks in medium- and low-resource languages. We trace this failure to semantic clustering, which fails when the tokenizer vocabulary contains too few full-word tokens for a given language. To address this, we introduce STEAM, a detection method that uses Bayesian optimisation to search among 126 candidate languages for the back-translation that best recovers the watermark strength. It is compatible with any watermarking method, robust across different tokenizers and languages, non-invasive, and easily extendable to new languages. With average gains of +0.23 AUC and +37%p TPR@1%, STEAM provides a scalable approach toward fairer watermarking across the diversity of languages.
Subjects:
Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
Cite as:
arXiv:2510.18019 [cs.CL]
(or
arXiv:2510.18019v3 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2510.18019
Focus to learn more
arXiv-issued DOI via DataCite
Submission history
From: Asim Mohamed [view email]
[v1]
Mon, 20 Oct 2025 18:51:20 UTC (899 KB)
[v2]
Wed, 25 Mar 2026 17:52:45 UTC (987 KB)
[v3]
Fri, 11 Sep 2026 17:40:56 UTC (1,060 KB)
Full-text links:
Access Paper:
View a PDF of the paper titled Is Multilingual LLM Watermarking Truly Multilingual? Scaling Robustness to 100+ Languages via Back-Translation, by Asim Mohamed and Martin Gubri
View PDF
HTML (experimental)
TeX Source
view license
Current browse context:
cs.CL
new
|
recent
| 2025-10
Change to browse by:
cs
cs.AI
References & Citations
NASA ADS
Google Scholar
Semantic Scholar
export BibTeX citation
Loading...
BibTeX formatted citation
×
loading...
Data provided by:
Bookmark
Bibliographic Tools
Bibliographic and Citation Tools
Bibliographic Explorer Toggle
Bibliographic Explorer (What is the Explorer?)
Connected Papers Toggle
Connected Papers (What is Connected Papers?)
Litmaps Toggle
Litmaps (What is Litmaps?)
scite.ai Toggle
scite Smart Citations (What are Smart Citations?)
Code, Data, Media
Code, Data and Media Associated with this Article
alphaXiv Toggle
alphaXiv (What is alphaXiv?)
Links to Code Toggle
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub Toggle
DagsHub (What is DagsHub?)
GotitPub Toggle
Gotit.pub (What is GotitPub?)
Huggingface Toggle
Hugging Face (What is Huggingface?)
ScienceCast Toggle
ScienceCast (What is ScienceCast?)
Demos
Demos
Replicate Toggle
Replicate (What is Replicate?)
Spaces Toggle
Hugging Face Spaces (What is Spaces?)
Spaces Toggle
TXYZ.AI (What is TXYZ.AI?)
Related Papers
Recommenders and Search Tools
Link to Influence Flower
Influence Flower (What are Influence Flowers?)
Core recommender toggle
CORE Recommender (What is CORE?)
Author
Venue
Institution
Topic
About arXivLabs
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Which authors of this paper are endorsers? |
Disable MathJax (What is MathJax?)