{
  "abstract": "Large Language Models (LLMs) frequently generate plausible but non-factual content, a phenomenon known as hallucination. While existing detection methods typically rely on computationally expensive sampling-based consistency checks or external knowledge retrieval, we propose a new method that treats the LLM as a black-box dynamical system. By projecting LLM responses into a high-dimensional manifold via an embedding model, we characterize the resulting vector sequences as observable realizations of the model's latent state-space dynamics. Leveraging Koopman operator theory, we fit the transition operators for both factual and hallucinated regimes and define a differential residual score based on their respective prediction errors. This approach enables low-cost hallucination detection in a single-sample pass, avoiding the need for secondary sampling or external grounding. Extensive testing across three data benchmarks demonstrates that our method achieves state-of-the-art performance with reduced resource overhead.",
  "abstract_shown": true,
  "affiliations": [],
  "article_type": null,
  "arxiv": "2605.05134",
  "author_total": 2,
  "authors": [
    {
      "name": "Dan Wilson"
    },
    {
      "name": "Mohamed Akrout"
    }
  ],
  "bibtex": "https://research.draconitic.com/paper/3d24n5tpre5tx.bib",
  "cited_by": 0,
  "doi": "10.48550/arxiv.2605.05134",
  "fulltext_url": null,
  "id": "3d24n5tpre5tx",
  "issns": [],
  "landing_url": "https://arxiv.org/abs/2605.05134",
  "language": null,
  "licence": "CC-BY-4.0",
  "licence_label": "CC BY 4.0",
  "licence_source": "article",
  "licence_url": "http://creativecommons.org/licenses/by/4.0/",
  "metadata_licence": "CC0-1.0",
  "native_subjects": [
    [
      "arxiv",
      "cs.LG"
    ],
    [
      "arxiv",
      "math.DS"
    ],
    [
      "openalex",
      "1702|Topic Modeling"
    ]
  ],
  "openalex": "W7160489409",
  "pdf_url": "https://arxiv.org/pdf/2605.05134",
  "pmcid": null,
  "pmid": null,
  "published": "2026-09-25",
  "publisher": null,
  "ris": "https://research.draconitic.com/paper/3d24n5tpre5tx.ris",
  "sources": [
    {
      "first_seen": "2026-10-30T12:24:57Z",
      "native_id": "2605.05134",
      "source": "arxiv",
      "url": "https://arxiv.org/abs/2605.05134"
    }
  ],
  "subjects": [
    {
      "field": "mathematics",
      "origin": "arxiv:math.DS",
      "subfield": "dynamical-systems"
    }
  ],
  "title": "Low-Cost Black-Box Detection of LLM Hallucinations via Dynamical System Prediction",
  "title_text": "Low-Cost Black-Box Detection of LLM Hallucinations via Dynamical System Prediction",
  "type": "preprint",
  "updated_at": "2026-10-30T12:24:57Z",
  "url": "https://research.draconitic.com/paper/3d24n5tpre5tx",
  "venue": "arXiv",
  "versions": []
}