@article{chen-etal-2025-repreguard,
    title = "{R}epre{G}uard: Detecting {LLM}-Generated Text by Revealing Hidden Representation Patterns",
    author = "Chen, Xin  and
      Wu, Junchao  and
      Yang, Shu  and
      Zhan, Runzhe  and
      Wu, Zeyu  and
      Luo, Ziyang  and
      Wang, Di  and
      Yang, Min  and
      Chao, Lidia S.  and
      Wong, Derek F.",
    journal = "Transactions of the Association for Computational Linguistics",
    volume = "13",
    year = "2025",
    address = "Cambridge, MA",
    publisher = "MIT Press",
    url = "https://preview.aclanthology.org/fix-opsupmap-display/2025.tacl-1.81/",
    doi = "10.1162/tacl.a.61",
    pages = "1812--1831",
    abstract = "Detecting content generated by large language models (LLMs) is crucial for preventing misuse and building trustworthy AI systems. Although existing detection methods perform well, their robustness in out-of-distribution (OOD) scenarios is still lacking. In this paper, we hypothesize that, compared to features used by existing detection methods, the internal representations of LLMs contain more comprehensive and raw features that can more effectively capture and distinguish the statistical pattern differences between LLM-generated texts (LGT) and human-written texts (HWT). We validated this hypothesis across different LLMs and observed significant differences in neural activation patterns when processing these two types of texts. Based on this, we propose RepreGuard, an efficient statistics-based detection method. Specifically, we first employ a surrogate model to collect representation of LGT and HWT, and extract the distinct activation feature that can better identify LGT. We can classify the text by calculating the projection score of the text representations along this feature direction and comparing with a precomputed threshold. Experimental results show that RepreGuard outperforms all baselines with average 94.92{\%} AUROC on both in-distribution and OOD scenarios, while also demonstrating robust resilience to various text sizes and mainstream attacks.1"
}