Temperature Matters: Enhancing Watermark Robustness Against Paraphrasing Attacks

TL;DR

This paper introduces a new watermarking technique for AI-generated text that remains effective against paraphrasing attacks, addressing concerns over misuse of large language models.

Contribution

It presents a novel watermarking method for synthetic text detection that demonstrates improved robustness against paraphrasing compared to existing techniques.

Findings

The proposed watermarking method outperforms previous approaches in robustness.

Experimental results confirm effectiveness against paraphrasing attacks.

The study emphasizes the importance of temperature settings in watermark robustness.

Abstract

In the present-day scenario, Large Language Models (LLMs) are establishing their presence as powerful instruments permeating various sectors of society. While their utility offers valuable support to individuals, there are multiple concerns over potential misuse. Consequently, some academic endeavors have sought to introduce watermarking techniques, characterized by the inclusion of markers within machine-generated text, to facilitate algorithmic identification. This research project is focused on the development of a novel methodology for the detection of synthetic text, with the overarching goal of ensuring the ethical application of LLMs in AI-driven text generation. The investigation commences with replicating findings from a previous baseline study, thereby underscoring its susceptibility to variations in the underlying generation model. Subsequently, we propose an innovative watermarking approach and subject it to rigorous evaluation, employing paraphrased generated text to asses its robustness. Experimental results highlight the robustness of our proposal compared to the~\cite{aarson} watermarking method.