多语言大模型水印技术是否真正具备多语言能力？一种简单的回译解决方案

摘要

多语言水印技术旨在使大型语言模型（LLM）的输出在不同语言间可追踪，然而现有方法仍显不足。尽管声称具备跨语言鲁棒性，这些方法仅在高资源语言上进行了评估。我们揭示，现有的多语言水印方法并非真正多语言：在中低资源语言下，面对翻译攻击时其鲁棒性丧失。我们将此失败归因于语义聚类，当分词器词汇表中针对某一语言的全词标记过少时，该机制失效。为解决这一问题，我们引入了STEAM，一种基于反向翻译的检测方法，能够恢复因翻译而削弱的水印强度。STEAM与任何水印方法兼容，在不同分词器和语言间均表现出鲁棒性，非侵入式设计，且易于扩展至新语言。在17种语言上，STEAM平均提升了+0.19的AUC值和+40%的TPR@1%，为跨语言公平水印提供了一条简单而稳健的路径。

English

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 back-translation-based detection method that restores watermark strength lost through translation. STEAM is compatible with any watermarking method, robust across different tokenizers and languages, non-invasive, and easily extendable to new languages. With average gains of +0.19 AUC and +40%p TPR@1% on 17 languages, STEAM provides a simple and robust path toward fairer watermarking across diverse languages.

多语言大模型水印技术是否真正具备多语言能力？一种简单的回译解决方案

Is Multilingual LLM Watermarking Truly Multilingual? A Simple Back-Translation Solution

摘要

Support