Abstract:Current techniques for detecting AI-generated text are largely confined to manual feature crafting and supervised binary classification paradigms. These methodologies typically lead to performance bottlenecks and unsatisfactory generalizability. Consequently, these methods are often inapplicable for out-of-distribution (OOD) data and newly emerged large language models (LLMs). In this paper, we revisit the task of AI-generated text detection. We argue that the key to accomplishing this task lies in distinguishing writing styles of different authors, rather than simply classifying the text into human-written or AI-generated text. To this end, we propose DeTeCtive, a multi-task auxiliary, multi-level contrastive learning framework. DeTeCtive is designed to facilitate the learning of distinct writing styles, combined with a dense information retrieval pipeline for AI-generated text detection. Our method is compatible with a range of text encoders. Extensive experiments demonstrate that our method enhances the ability of various text encoders in detecting AI-generated text across multiple benchmarks and achieves state-of-the-art results. Notably, in OOD zero-shot evaluation, our method outperforms existing approaches by a large margin. Moreover, we find our method boasts a Training-Free Incremental Adaptation (TFIA) capability towards OOD data, further enhancing its efficacy in OOD detection scenarios. We will open-source our code and models in hopes that our work will spark new thoughts in the field of AI-generated text detection, ensuring safe application of LLMs and enhancing compliance. Our code is available at <a class="link-external link-https" href="https://github.com/heyongxin233/DeTeCtive" rel="external noopener nofollow">this https URL</a>.

A Novel Textual Data Augmentation Method for Identifying Comparative Text from User-Generated Content

A review for comparative text mining: From data acquisition to practical application

Diversified text-to-image generation via deep mutual information estimation

Learning Extraction of Chinese Comparative Sentences for Evaluative Text

Exploiting Machine Learning for Comparative Sentences Extraction

Fusing Various Document Representations for Comparative Text Identification from Product Reviews

Chinese Comparative Sentence Identification Based on the Combination of Rules and Statistics

DeTeCtive: Detecting AI-generated Text via Multi-Level Contrastive Learning

Automatic Identify Chinese Comparative Sentences

Recognizing Comparative Sentences from Chinese Review Texts

Comparable Entity Mining from Comparative Questions

A Comparative Analysis of Text Mining Methodologies for Online Consumer Reviews

Comparison of Quality Indicators in User-generated Content Using Social Media and Scholarly Text

Utilizing Non-Parallel Text for Style Transfer by Making Partial Comparisons.

A Deep Learning Based Method to Measure the Similarity of Long Text

A Survey On Text-to-3D Contents Generation In The Wild

Improving embedding with contrastive fine-tuning on small datasets with expert-augmented scores

STA: Self-controlled Text Augmentation for Improving Text Classifications

Aspect Term Extraction via Contrastive Learning over Self-augmented Data

Mutual Clustering on Comparative Texts via Heterogeneous Information Networks

LALDM: A Multimodal Aspect Level Text Analysis Method and Its Application in Online Consumer Electronics