MGT-Prism: Enhancing Domain Generalization for Machine-Generated Text Detection via Spectral Alignment

TL;DR

MGT-Prism introduces a spectral domain approach for detecting machine-generated texts, significantly improving cross-domain detection accuracy by filtering domain-sensitive features and aligning spectral patterns.

Contribution

The paper proposes a novel frequency domain spectral alignment method to enhance domain generalization in machine-generated text detection.

Findings

Outperforms state-of-the-art methods by 0.90% in accuracy

Achieves 0.92% higher F1 score across multiple datasets

Effective in three different domain-generalization scenarios

Abstract

Large Language Models have shown growing ability to generate fluent and coherent texts that are highly similar to the writing style of humans. Current detectors for Machine-Generated Text (MGT) perform well when they are trained and tested in the same domain but generalize poorly to unseen domains, due to domain shift between data from different sources. In this work, we propose MGT-Prism, an MGT detection method from the perspective of the frequency domain for better domain generalization. Our key insight stems from analyzing text representations in the frequency domain, where we observe consistent spectral patterns across diverse domains, while significant discrepancies in magnitude emerge between MGT and human-written texts (HWTs). The observation initiates the design of a low frequency domain filtering module for filtering out the document-level features that are sensitive to domain shift, and a dynamic spectrum alignment strategy to extract the task-specific and domain-invariant features for improving the detector's performance in domain generalization. Extensive experiments demonstrate that MGT-Prism outperforms state-of-the-art baselines by an average of 0.90% in accuracy and 0.92% in F1 score on 11 test datasets across three domain-generalization scenarios.