English

Multi-turn Physics-informed Vision-language Model for Physics-grounded Anomaly Detection

Computer Vision and Pattern Recognition 2026-03-17 v1

Abstract

Vision-Language Models (VLMs) demonstrate strong general-purpose reasoning but remain limited in physics-grounded anomaly detection, where causal understanding of dynamics is essential. Existing VLMs, trained predominantly on appearance-centric correlations, fail to capture kinematic constraints, leading to poor performance on anomalies such as irregular rotations or violated mechanical motions. We introduce a physics-informed instruction tuning framework that explicitly encodes object properties, motion paradigms, and dynamic constraints into structured prompts. By delivering these physical priors through multi-turn dialogues, our method decomposes causal reasoning into incremental steps, enabling robust internal representations of normal and abnormal dynamics. Evaluated on the Phys-AD benchmark, our approach achieves 96.7% AUROC in video-level detection--substantially outperforming prior SOTA (66.9%)--and yields superior causal explanations (0.777 LLM score). This work highlights how structured physics priors can transform VLMs into reliable detectors of dynamic anomalies.

Keywords

Cite

@article{arxiv.2603.15237,
  title  = {Multi-turn Physics-informed Vision-language Model for Physics-grounded Anomaly Detection},
  author = {Yao Gu and Xiaohao Xu and Yingna Wu},
  journal= {arXiv preprint arXiv:2603.15237},
  year   = {2026}
}

Comments

Accepted by IEEE ICASSP2026

R2 v1 2026-07-01T11:22:13.816Z