Detecting Subtle Differences between Human and Model Languages Using Spectrum of Relative Likelihood

TL;DR

This paper introduces a novel approach using the spectrum of relative likelihoods to detect subtle differences between human and AI-generated texts, achieving state-of-the-art results and providing insights rooted in psycholinguistics.

Contribution

It presents a new detection method based on relative likelihood spectrum analysis, outperforming previous zero-shot techniques and revealing subtle language differences with theoretical backing.

Findings

Achieves state-of-the-art short-text detection performance

Provides a new perspective using relative likelihood spectrum analysis

Reveals subtle linguistic differences rooted in psycholinguistics

Abstract

Human and model-generated texts can be distinguished by examining the magnitude of likelihood in language. However, it is becoming increasingly difficult as language model's capabilities of generating human-like texts keep evolving. This study provides a new perspective by using the relative likelihood values instead of absolute ones, and extracting useful features from the spectrum-view of likelihood for the human-model text detection task. We propose a detection procedure with two classification methods, supervised and heuristic-based, respectively, which results in competitive performances with previous zero-shot detection methods and a new state-of-the-art on short-text detection. Our method can also reveal subtle differences between human and model languages, which find theoretical roots in psycholinguistics studies. Our code is available at https://github.com/CLCS-SUSTech/FourierGPT