Less is More: Sparse Watermarking in LLMs with Enhanced Text Quality

TL;DR

This paper introduces Sparse Watermarking for LLMs, which applies watermarks to select tokens based on POS tags, balancing detection effectiveness with high-quality text generation.

Contribution

It proposes a novel sparse watermarking method that improves text quality while maintaining high detectability compared to existing watermarking techniques.

Findings

High detectability of watermarks across tasks

Improved text quality over previous methods

Effective distribution of watermarked tokens

Abstract

With the widespread adoption of Large Language Models (LLMs), concerns about potential misuse have emerged. To this end, watermarking has been adapted to LLM, enabling a simple and effective way to detect and monitor generated text. However, while the existing methods can differentiate between watermarked and unwatermarked text with high accuracy, they often face a trade-off between the quality of the generated text and the effectiveness of the watermarking process. In this work, we present a novel type of LLM watermark, Sparse Watermark, which aims to mitigate this trade-off by applying watermarks to a small subset of generated tokens distributed across the text. The key strategy involves anchoring watermarked tokens to words that have specific Part-of-Speech (POS) tags. Our experimental results demonstrate that the proposed watermarking scheme achieves high detectability while generating text that outperforms previous LLM watermarking methods in quality across various tasks