Watermarking Language Models with Error Correcting Codes

TL;DR

This paper introduces a novel watermarking method for large language models using error correcting codes, which is robust, undetectable to humans, and maintains output quality.

Contribution

It proposes the RBC watermarking framework that encodes signals with error correcting codes, providing robustness and theoretical guarantees while preserving text quality.

Findings

Watermark is robust to edits, deletions, and translations.

The method maintains high output quality without degradation.

Empirical results show it outperforms existing watermarking techniques.

Abstract

Recent progress in large language models enables the creation of realistic machine-generated content. Watermarking is a promising approach to distinguish machine-generated text from human text, embedding statistical signals in the output that are ideally undetectable to humans. We propose a watermarking framework that encodes such signals through an error correcting code. Our method, termed robust binary code (RBC) watermark, introduces no noticeable degradation in quality. We evaluate our watermark on base and instruction fine-tuned models and find that our watermark is robust to edits, deletions, and translations. We provide an information-theoretic perspective on watermarking, a powerful statistical test for detection and for generating $p$-values, and theoretical guarantees. Our empirical findings suggest our watermark is fast, powerful, and robust, comparing favorably to the state-of-the-art.