Saved in:
| Main Authors: | , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2603.06081 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866914375006158848 |
|---|---|
| author | Luan, Bozhi Li, Gen Qin, Yalan Guo, Jifeng Zhou, Yun Wu, Faguo Zheng, Hongwei Wu, Wenjun Fan, Zhaoxin |
| author_facet | Luan, Bozhi Li, Gen Qin, Yalan Guo, Jifeng Zhou, Yun Wu, Faguo Zheng, Hongwei Wu, Wenjun Fan, Zhaoxin |
| contents | We address hallucination detection in Large Language Models (LLMs) and Multimodal Large Language Models (MLLMs) by framing the problem through the lens of dynamical systems stability theory. Rather than treating hallucination as a straightforward classification task, we conceptualize (M)LLMs as dynamical systems, where factual knowledge is represented by stable equilibrium points within the representation space. Our main insight is that hallucinations tend to arise at the boundaries of knowledge-transition regions separating stable and unstable zones. To capture this phenomenon, we propose Lyapunov Probes: lightweight networks trained with derivative-based stability constraints that enforce a monotonic decay in confidence under input perturbations. By performing systematic perturbation analysis and applying a two-stage training process, these probes reliably distinguish between stable factual regions and unstable, hallucination-prone regions. Experiments on diverse datasets and models demonstrate consistent improvements over existing baselines. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_06081 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Lyapunov Probes for Hallucination Detection in Large Foundation Models Luan, Bozhi Li, Gen Qin, Yalan Guo, Jifeng Zhou, Yun Wu, Faguo Zheng, Hongwei Wu, Wenjun Fan, Zhaoxin Computer Vision and Pattern Recognition We address hallucination detection in Large Language Models (LLMs) and Multimodal Large Language Models (MLLMs) by framing the problem through the lens of dynamical systems stability theory. Rather than treating hallucination as a straightforward classification task, we conceptualize (M)LLMs as dynamical systems, where factual knowledge is represented by stable equilibrium points within the representation space. Our main insight is that hallucinations tend to arise at the boundaries of knowledge-transition regions separating stable and unstable zones. To capture this phenomenon, we propose Lyapunov Probes: lightweight networks trained with derivative-based stability constraints that enforce a monotonic decay in confidence under input perturbations. By performing systematic perturbation analysis and applying a two-stage training process, these probes reliably distinguish between stable factual regions and unstable, hallucination-prone regions. Experiments on diverse datasets and models demonstrate consistent improvements over existing baselines. |
| title | Lyapunov Probes for Hallucination Detection in Large Foundation Models |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2603.06081 |