Does DetectGPT Fully Utilize Perturbation? Bridging Selective Perturbation to Fine-tuned Contrastive Learning Detector would be Better

TL;DR

This paper introduces Pecola, a fine-tuned contrastive learning detector that improves upon DetectGPT by selectively perturbing tokens, leading to better accuracy and robustness in detecting machine-generated text.

Contribution

It bridges metric-based and fine-tuned detection methods through selective perturbation and contrastive learning, enhancing detection performance and generalization.

Findings

Pecola outperforms SOTA by 1.20% in accuracy on four datasets.

Selective perturbation improves detection robustness.

Contrastive learning enhances generalization in detection.

Abstract

The burgeoning generative capabilities of large language models (LLMs) have raised growing concerns about abuse, demanding automatic machine-generated text detectors. DetectGPT, a zero-shot metric-based detector, first introduces perturbation and shows great performance improvement. However, in DetectGPT, the random perturbation strategy could introduce noise, and logit regression depends on the threshold, harming the generalizability and applicability of individual or small-batch inputs. Hence, we propose a novel fine-tuned detector, Pecola, bridging metric-based and fine-tuned methods by contrastive learning on selective perturbation. Selective strategy retains important tokens during perturbation and weights for multi-pair contrastive learning. The experiments show that Pecola outperforms the state-of-the-art (SOTA) by 1.20% in accuracy on average on four public datasets. And we further analyze the effectiveness, robustness, and generalization of the method.