@inproceedings{mohammadzadeh-etal-2025-hallucination,
    title = "Hallucination Detox: Sensitivity Dropout ({S}en{D}) for Large Language Model Training",
    author = "Mohammadzadeh, Shahrad  and
      Guerra, Juan David  and
      Bonizzato, Marco  and
      Rabbany, Reihaneh  and
      Farnadi, Golnoosh",
    editor = "Che, Wanxiang  and
      Nabende, Joyce  and
      Shutova, Ekaterina  and
      Pilehvar, Mohammad Taher",
    booktitle = "Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
    month = jul,
    year = "2025",
    address = "Vienna, Austria",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/ingestion-acl-25/2025.acl-long.276/",
    pages = "5538--5554",
    ISBN = "979-8-89176-251-0",
    abstract = "As large language models (LLMs) become increasingly prevalent, concerns about their reliability, particularly due to hallucinations - factually inaccurate or irrelevant outputs - have grown. Our research investigates the relationship between the uncertainty in training dynamics and the emergence of hallucinations. Using models from the Pythia suite and several hallucination detection metrics, we analyze hallucination trends and identify significant variance during training. To address this, we propose Sensitivity Dropout (SenD), a novel training protocol designed to reduce hallucination variance during training by deterministically dropping embedding indices with significant variability. In addition, we develop an unsupervised hallucination detection metric, Efficient EigenScore (EES), which approximates the traditional EigenScore in 2x speed. This metric is integrated into our training protocol, allowing SenD to be both computationally scalable and effective at reducing hallucination variance. SenD improves test-time reliability of Pythia and Meta{'}s Llama models by up to 17{\%} and enhances factual accuracy in Wikipedia, Medical, Legal, and Coding domains without affecting downstream task performance."
}