🤖 AI Summary
Existing LLM-generated text detection methods suffer from poor generalizability, vulnerability to paraphrasing, limited interpretability, and reliance on access to target models or labeled data.
Method: This paper proposes an unsupervised, cross-domain dual-track stylistic analysis framework that jointly models discrete stylistic features (e.g., syntactic and lexical patterns) and continuous semantic embeddings. It employs multi-domain style alignment and an end-to-end unsupervised classifier for unified representation learning—requiring neither model internals nor human annotations.
Contribution/Results: The method achieves state-of-the-art in-domain accuracy across five diverse domains; improves cross-domain detection performance by up to 36.30% over baselines; and demonstrates strong robustness against human-AI hybrid texts and adversarial paraphrasing. These results empirically validate stylistic signals as reliable, generalizable indicators for trustworthy detection.
📝 Abstract
With the increasing integration of large language models (LLMs) into open-domain writing, detecting machine-generated text has become a critical task for ensuring content authenticity and trust. Existing approaches rely on statistical discrepancies or model-specific heuristics to distinguish between LLM-generated and human-written text. However, these methods struggle in real-world scenarios due to limited generalization, vulnerability to paraphrasing, and lack of explainability, particularly when facing stylistic diversity or hybrid human-AI authorship. In this work, we propose StyleDecipher, a robust and explainable detection framework that revisits LLM-generated text detection using combined feature extractors to quantify stylistic differences. By jointly modeling discrete stylistic indicators and continuous stylistic representations derived from semantic embeddings, StyleDecipher captures distinctive style-level divergences between human and LLM outputs within a unified representation space. This framework enables accurate, explainable, and domain-agnostic detection without requiring access to model internals or labeled segments. Extensive experiments across five diverse domains, including news, code, essays, reviews, and academic abstracts, demonstrate that StyleDecipher consistently achieves state-of-the-art in-domain accuracy. Moreover, in cross-domain evaluations, it surpasses existing baselines by up to 36.30%, while maintaining robustness against adversarial perturbations and mixed human-AI content. Further qualitative and quantitative analysis confirms that stylistic signals provide explainable evidence for distinguishing machine-generated text. Our source code can be accessed at https://github.com/SiyuanLi00/StyleDecipher.