SimMark通过语义相似性嵌入水印，实现对大模型文本的隐蔽追踪。
SimMark: A Robust Sentence-Level Similarity-Based Watermarking Algorithm for Large Language Models
- 利用句子嵌入相似性与拒绝采样，嵌入人类不可察觉的统计模式。
- 在多种改写攻击下仍保持高检测率，优于现有句级水印方法。
- 无需模型内部信息，适用于开源和API模型，适合内容溯源场景。
大语言模型（LLMs）的广泛应用亟需可靠的生成文本检测方法。我们提出SimMark，一种无需访问模型内部信息的鲁棒句级水印算法，兼容开源与API形式的LLMs。该方法结合语义句子嵌入的相似性与拒绝采样，嵌入人类不可察觉的可检测统计模式，并采用软计数机制，有效抵御改写攻击。实验表明，SimMark在鲁棒性、采样效率和跨领域适用性方面均超越现有句级水印技术，同时保持文本质量与流畅性，树立了新的基准。
原文摘要 · Abstract (English)
The widespread adoption of large language models (LLMs) necessitates reliable methods to detect LLM-generated text. We introduce SimMark, a robust sentence-level watermarking algorithm that makes LLMs' outputs traceable without requiring access to model internals, making it compatible with both open and API-based LLMs. By leveraging the similarity of semantic sentence embeddings combined with rejection sampling to embed detectable statistical patterns imperceptible to humans, and employing a soft counting mechanism, SimMark achieves robustness against paraphrasing attacks. Experimental results demonstrate that SimMark sets a new benchmark for robust watermarking of LLM-generated content, surpassing prior sentence-level watermarking techniques in robustness, sampling efficiency, and applicability across diverse domains, all while maintaining the text quality and fluency.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。