@inproceedings{zhang-etal-2025-law,
    title = "The Law of Knowledge Overshadowing: Towards Understanding, Predicting, and Preventing {LLM} Hallucination",
    author = "Zhang, Yuji  and
      Li, Sha  and
      Qian, Cheng  and
      Liu, Jiateng  and
      Yu, Pengfei  and
      Han, Chi  and
      Fung, Yi R.  and
      McKeown, Kathleen  and
      Zhai, ChengXiang  and
      Li, Manling  and
      Ji, Heng",
    editor = "Akhtar, Mubashara  and
      Aly, Rami  and
      Christodoulopoulos, Christos  and
      Cocarascu, Oana  and
      Guo, Zhijiang  and
      Mittal, Arpit  and
      Schlichtkrull, Michael  and
      Thorne, James  and
      Vlachos, Andreas",
    booktitle = "Proceedings of the Eighth Fact Extraction and VERification Workshop (FEVER)",
    month = jul,
    year = "2025",
    address = "Vienna, Austria",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/acl25-workshop-ingestion/2025.fever-1.10/",
    pages = "132--150",
    ISBN = "978-1-959429-53-1",
    abstract = "Hallucination is a persistent challenge in large language models (LLMs), where even with rigorous quality control, models often generate distorted facts. This paradox, in which error generation continues despite high-quality training data, calls for a deeper understanding of the underlying LLM mechanisms. To address it, we propose a novel concept: knowledge overshadowing, where model{'}s dominant knowledge can obscure less prominent knowledge during text generation, causing the model to fabricate inaccurate details. Building on this idea, we introduce a novel framework to quantify factual hallucinations by modeling knowledge overshadowing. Central to our approach is the log-linear law, which predicts that the rate of factual hallucination increases linearly with the logarithmic scale of (1) Knowledge Popularity, (2) Knowledge Length, and (3) Model Size. The law provides a means to preemptively quantify hallucinations, offering foresight into their occurrence even before model training or inference. Built on the overshadowing effect, we propose a new decoding strategy CoDA, to mitigate hallucinations, which notably enhances model factuality on Overshadow (27.9{\%}), MemoTrap (13.1{\%}) and NQ-Swap (18.3{\%}). Our findings not only deepen understandings of the underlying mechanisms behind hallucinations but also provide actionable insights for developing more predictable and controllable language models."
}