@inproceedings{gautam-etal-2026-energy,
    title = "The Energy of Falsehood: Detecting Hallucinations via Diffusion Model Likelihoods",
    author = "Gautam, Arpit Singh  and
      Talreja, Kailash  and
      Jha, Saurabh",
    editor = "Akhtar, Mubashara  and
      Aly, Rami  and
      Cao, Rui  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 Ninth Fact Extraction and {VER}ification Workshop ({FEVER})",
    month = mar,
    year = "2026",
    address = "Rabat, Morocco",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/credits/2026.fever-1.4/",
    pages = "47--58",
    ISBN = "979-8-89176-365-4",
    abstract = "Large Language Models (LLMs) frequently ``hallucinate'' plausible but incorrect assertions, a vulnerability often missed by uncertainty metrics when models are ``confidently wrong.'' We propose DiffuTruth, an unsupervised framework that re-conceptualizes fact verification via non-equilibrium thermodynamics, positing that factual truths act as stable attractors on a generative manifold while hallucinations are unstable. We introduce the ``Generative Stress Test'': claims are corrupted with noise and reconstructed using a discrete text diffusion model. We define Semantic Energy, a metric measuring the semantic divergence between the original claim and its reconstruction using an NLI critic. Unlike vector-space errors, Semantic Energy isolates deep factual contradictions. We further propose a Hybrid Calibration fusing this stability signal with discriminative confidence. Extensive experiments on FEVER demonstrate DiffuTruth achieves a state-of-the-art unsupervised AUROC of 0.725, outperforming baselines by +1.5{\%} through the correction of overconfident predictions. Furthermore, we show superior zero-shot generalization on the multi-hop HOVER dataset, outperforming baselines by over 4{\%}, confirming the robustness of thermodynamic truth properties to distribution shifts."
}