Beyond Raw Detection Scores: Markov-Informed Calibration for Boosting Machine-Generated Text Detection

TL;DR

This paper introduces a Markov-informed calibration method for machine-generated text detection that improves accuracy by modeling token-level score biases, with minimal computational cost.

Contribution

It proposes a novel calibration strategy using Markov random fields to address token score biases in metric-based detection methods, enhancing detection robustness.

Findings

Significant performance improvements over baseline detectors.

Effective in cross-LLM and paraphrasing attack scenarios.

Minimal additional computational overhead.

Abstract

While machine-generated texts (MGTs) offer great convenience, they also pose risks such as disinformation and phishing, highlighting the need for reliable detection. Metric-based methods, which extract statistically distinguishable features of MGTs, are often more practical than complex model-based methods that are prone to overfitting. Given their diverse designs, we first place representative metric-based methods within a unified framework, enabling a clear assessment of their advantages and limitations. Our analysis identifies a core challenge across these methods: the token-level detection score is easily biased by the inherent randomness of the MGTs generation process. To address this, we theoretically and empirically reveal two relationships of context detection scores that may aid calibration: Neighbor Similarity and Initial Instability. We then propose a Markov-informed score calibration strategy that models these relationships using Markov random fields, and implements it as a lightweight component via a mean-field approximation, allowing our method to be seamlessly integrated into existing detectors. Extensive experiments in various real-world scenarios, such as cross-LLM and paraphrasing attacks, demonstrate significant gains over baselines with negligible computational overhead. The code is available at https://github.com/tmlr-group/MRF_Calibration.