无需模型内部信息即可检测大模型输出，且不影响生成质量。
A Watermark for Black-Box Language Models
- 仅需调用API采样即可实现水印嵌入，适用于黑盒场景。
- 水印无失真，可使用多个密钥嵌套或串联，提升安全性。
- 在无法获取概率分布时仍优于传统白盒方案，适合实际部署。
水印技术已成为检测大语言模型输出的有效手段。然而，现有方法大多需要白盒访问模型的下一个词概率分布，这对LLM API的下游用户通常不可行。本文提出一种基于黑盒采样的原理性水印方案，仅需通过API生成序列即可完成水印嵌入，具备无失真特性，并支持多密钥嵌套与链式使用。我们提供了性能保证，展示了其在白盒场景下的扩展能力，并通过全面实验表明，在缺乏白盒访问时该方案仍可超越现有白盒方法。
原文摘要 · Abstract (English)
Watermarking has recently emerged as an effective strategy for detecting the outputs of large language models (LLMs). Most existing schemes require white-box access to the model's next-token probability distribution, which is typically not accessible to downstream users of an LLM API. In this work, we propose a principled watermarking scheme that requires only the ability to sample sequences from the LLM (i.e. black-box access), boasts a distortion-free property, and can be chained or nested using multiple secret keys. We provide performance guarantees, demonstrate how it can be leveraged when white-box access is available, and show when it can outperform existing white-box schemes via comprehensive experiments.
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。