Is Contrasting All You Need? Contrastive Learning for the Detection and Attribution of AI-generated Text

TL;DR

This paper introduces WhosAI, a contrastive learning framework that effectively detects and attributes AI-generated text by learning semantic similarities across multiple generators, outperforming existing methods on benchmark datasets.

Contribution

The paper presents a novel triplet-network contrastive learning approach that handles both detection and attribution tasks simultaneously and is scalable to new AI text-generation models.

Findings

Outperforms all methods on TuringBench benchmark

Effectively handles multiple AI generators simultaneously

Achieves state-of-the-art results in detection and attribution

Abstract

The significant progress in the development of Large Language Models has contributed to blurring the distinction between human and AI-generated text. The increasing pervasiveness of AI-generated text and the difficulty in detecting it poses new challenges for our society. In this paper, we tackle the problem of detecting and attributing AI-generated text by proposing WhosAI, a triplet-network contrastive learning framework designed to predict whether a given input text has been generated by humans or AI and to unveil the authorship of the text. Unlike most existing approaches, our proposed framework is conceived to learn semantic similarity representations from multiple generators at once, thus equally handling both detection and attribution tasks. Furthermore, WhosAI is model-agnostic and scalable to the release of new AI text-generation models by incorporating their generated instances into the embedding space learned by our framework. Experimental results on the TuringBench benchmark of 200K news articles show that our proposed framework achieves outstanding results in both the Turing Test and Authorship Attribution tasks, outperforming all the methods listed in the TuringBench benchmark leaderboards.