从权衡到协同：面向大语言模型的多功能共生水印框架

From Trade-off to Synergy: A Versatile Symbiotic Watermarking Framework for Large Language Models

May 15, 2025

作者: Yidan Wang, Yubing Ren, Yanan Cao, Binxing Fang

cs.AI

摘要

大型语言模型（LLMs）的兴起加剧了人们对AI生成文本滥用的担忧，使得水印技术成为一种颇具前景的解决方案。当前主流的LLM水印方案主要分为两类：基于logits的和基于采样的。然而，现有方案在鲁棒性、文本质量和安全性之间往往存在权衡。为缓解这一问题，我们整合了基于logits和基于采样的方案，充分发挥各自优势以实现协同效应。本文中，我们提出了一种多功能共生水印框架，包含三种策略：串行、并行及混合。该混合框架通过利用令牌熵和语义熵自适应地嵌入水印，优化了可检测性、鲁棒性、文本质量与安全性之间的平衡。此外，我们在多种数据集和模型上进行了全面实验以验证方法的有效性。实验结果表明，我们的方法超越了现有基线，达到了当前最优（SOTA）性能。我们相信这一框架为多样化的水印范式提供了新的见解。代码已开源，详见https://github.com/redwyd/SymMark{https://github.com/redwyd/SymMark}。

English

The rise of Large Language Models (LLMs) has heightened concerns about the misuse of AI-generated text, making watermarking a promising solution. Mainstream watermarking schemes for LLMs fall into two categories: logits-based and sampling-based. However, current schemes entail trade-offs among robustness, text quality, and security. To mitigate this, we integrate logits-based and sampling-based schemes, harnessing their respective strengths to achieve synergy. In this paper, we propose a versatile symbiotic watermarking framework with three strategies: serial, parallel, and hybrid. The hybrid framework adaptively embeds watermarks using token entropy and semantic entropy, optimizing the balance between detectability, robustness, text quality, and security. Furthermore, we validate our approach through comprehensive experiments on various datasets and models. Experimental results indicate that our method outperforms existing baselines and achieves state-of-the-art (SOTA) performance. We believe this framework provides novel insights into diverse watermarking paradigms. Our code is available at https://github.com/redwyd/SymMark{https://github.com/redwyd/SymMark}.