@inproceedings{bazarova-etal-2026-hallucination,
    title = "Hallucination Detection in {LLM}s with Topological Divergence on Attention Graphs",
    author = "Bazarova, Alexandra  and
      Volodichev, Andrei  and
      Yugay, Aleksandr  and
      Shulga, Andrey  and
      Ermilova, Alina  and
      Polev, Konstantin  and
      Belikova, Julia  and
      Parchiev, Rauf  and
      Simakov, Dmitry  and
      Savchenko, Maxim  and
      Savchenko, Andrey  and
      Barannikov, Serguei  and
      Zaytsev, Alexey",
    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.704/",
    pages = "15449--15470",
    ISBN = "979-8-89176-390-6",
    abstract = "Hallucinations remain a critical challenge for large language models (LLMs), particularly in Retrieval-Augmented Generation (RAG) settings where models may generate outputs unsupported by the provided context. To address this, we introduce TOHA, a TOpology-based HAllucination detector, which leverages a topological divergence metric to quantify the structural properties of graphs induced by attention matrices. Examining the topological divergence between prompt and response subgraphs in RAG settings reveals consistent patterns: higher divergence values in specific attention heads correlate with unfaithful outputs, independent of the dataset. Extensive experiments {---} including evaluations on question answering and summarization tasks {---} show that our approach achieves state-of-the-art or competitive results on several benchmarks while requiring minimal annotated data and computational resources. Our findings indicate that the topological structure of attention matrices provides an efficient and robust metric for assessing the correctness of LLM{'}s responses."
}