@inproceedings{luo-etal-2026-two,
    title = "Two Pathways to Truthfulness: On the Intrinsic Encoding of {LLM} Hallucinations",
    author = "Luo, Wen  and
      Peng, Guangyue  and
      Li, Wei  and
      Wei, Shaohang  and
      Song, Feifan  and
      Wang, Liang  and
      Yang, Nan  and
      Zhang, Xingxing  and
      Jin, Jing  and
      Wei, Furu  and
      Wang, Houfeng",
    editor = "Liakata, Maria  and
      Moreira, Viviane P.  and
      Zhang, Jiajun  and
      Jurgens, David",
    booktitle = "Proceedings of the 64th Annual Meeting of the {A}ssociation for {C}omputational {L}inguistics (Volume 1: Long Papers)",
    month = jul,
    year = "2026",
    address = "San Diego, California, United States",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/ingest-acl/2026.acl-long.1173/",
    pages = "25580--25626",
    ISBN = "979-8-89176-390-6",
    abstract = "Despite their impressive capabilities, large language models (LLMs) frequently generate hallucinations. Previous work shows that their internal states encode rich signals of truthfulness, yet the origins and mechanisms of these signals remain unclear. In this paper, we demonstrate that truthfulness cues arise from two distinct information pathways: (1) a Question-Anchored pathway that depends on question{--}answer information flow, and (2) an Answer-Anchored pathway that derives self-contained evidence from the generated answer itself. First, we validate and disentangle these pathways through attention knockout and token patching. Afterwards, we uncover notable and intriguing properties of these two mechanisms. Further experiments reveal that (1) the two mechanisms are closely associated with LLM knowledge boundaries; and (2) internal representations are aware of their distinctions. Finally, building on these insightful findings, two applications are proposed to enhance hallucination detection performance. Overall, our work provides new insight into how LLMs internally encode truthfulness, offering directions for more reliable and self-aware generative systems."
}