Saved in:
Bibliographic Details
Main Authors: Chen, Liang, Bian, Yatao, Deng, Yang, Cai, Deng, Li, Shuaiyi, Zhao, Peilin, Wong, Kam-fai
Format: Preprint
Published: 2023
Subjects:
Online Access:https://arxiv.org/abs/2311.09832
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914826217848832
author Chen, Liang
Bian, Yatao
Deng, Yang
Cai, Deng
Li, Shuaiyi
Zhao, Peilin
Wong, Kam-fai
author_facet Chen, Liang
Bian, Yatao
Deng, Yang
Cai, Deng
Li, Shuaiyi
Zhao, Peilin
Wong, Kam-fai
contents Text watermarking has emerged as a pivotal technique for identifying machine-generated text. However, existing methods often rely on arbitrary vocabulary partitioning during decoding to embed watermarks, which compromises the availability of suitable tokens and significantly degrades the quality of responses. This study assesses the impact of watermarking on different capabilities of large language models (LLMs) from a cognitive science lens. Our finding highlights a significant disparity; knowledge recall and logical reasoning are more adversely affected than language generation. These results suggest a more profound effect of watermarking on LLMs than previously understood. To address these challenges, we introduce Watermarking with Mutual Exclusion (WatME), a novel approach leveraging linguistic prior knowledge of inherent lexical redundancy in LLM vocabularies to seamlessly integrate watermarks. Specifically, WatME dynamically optimizes token usage during the decoding process by applying a mutually exclusive rule to the identified lexical redundancies. This strategy effectively prevents the unavailability of appropriate tokens and preserves the expressive power of LLMs. We provide both theoretical analysis and empirical evidence showing that WatME effectively preserves the diverse capabilities of LLMs while ensuring watermark detectability.
format Preprint
id arxiv_https___arxiv_org_abs_2311_09832
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle WatME: Towards Lossless Watermarking Through Lexical Redundancy
Chen, Liang
Bian, Yatao
Deng, Yang
Cai, Deng
Li, Shuaiyi
Zhao, Peilin
Wong, Kam-fai
Computation and Language
Text watermarking has emerged as a pivotal technique for identifying machine-generated text. However, existing methods often rely on arbitrary vocabulary partitioning during decoding to embed watermarks, which compromises the availability of suitable tokens and significantly degrades the quality of responses. This study assesses the impact of watermarking on different capabilities of large language models (LLMs) from a cognitive science lens. Our finding highlights a significant disparity; knowledge recall and logical reasoning are more adversely affected than language generation. These results suggest a more profound effect of watermarking on LLMs than previously understood. To address these challenges, we introduce Watermarking with Mutual Exclusion (WatME), a novel approach leveraging linguistic prior knowledge of inherent lexical redundancy in LLM vocabularies to seamlessly integrate watermarks. Specifically, WatME dynamically optimizes token usage during the decoding process by applying a mutually exclusive rule to the identified lexical redundancies. This strategy effectively prevents the unavailability of appropriate tokens and preserves the expressive power of LLMs. We provide both theoretical analysis and empirical evidence showing that WatME effectively preserves the diverse capabilities of LLMs while ensuring watermark detectability.
title WatME: Towards Lossless Watermarking Through Lexical Redundancy
topic Computation and Language
url https://arxiv.org/abs/2311.09832