Robust AI-Generated Text Detection by Restricted Embeddings
Abstract: Growing amount and quality of AI-generated texts makes detecting such content more difficult. In most real-world scenarios, the domain (style and topic) of generated data and the generator model are not known in advance. In this work, we focus on the robustness of classifier-based detectors of AI-generated text, namely their ability to transfer to unseen generators or semantic domains. We investigate the geometry of the embedding space of Transformer-based text encoders and show that clearing out harmful linear subspaces helps to train a robust classifier, ignoring domain-specific spurious features. We investigate several subspace decomposition and feature selection strategies and achieve significant improvements over state of the art methods in cross-domain and cross-generator transfer. Our best approaches for head-wise and coordinate-based subspace removal increase the mean out-of-distribution (OOD) classification score by 9% and 14% for RoBERTa and BERT embeddings respectively.
Paper Type: Long
Research Area: Machine Learning for NLP
Research Area Keywords: analysis, interpretability, robustness, artificial text detection
Contribution Types: Model analysis & interpretability, NLP engineering experiment
Languages Studied: English
Submission Number: 5291
Loading