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. To accommodate varying user requirements and domain-specific sensitivities, we introduce a preference-aware calibration mechanism that optimizes the classification threshold based on a small set of demonstrations. 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.
| Subjects: | Machine Learning (cs.LG); Dynamical Systems (math.DS) |
| Cite as: | arXiv:2605.05134 [cs.LG] |
| (or arXiv:2605.05134v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2605.05134 arXiv-issued DOI via DataCite (pending registration) |
From: Mohamed Akrout [view email]
[v1]
Wed, 6 May 2026 17:07:29 UTC (1,369 KB)
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。