Physics Attention: A Physics-Learned Generative World Model with Disentangled Latent Physical Factors for Fluid
Open Access
- 9 November 2026
- conference paper
- Published by Association for Computing Machinery (ACM)
- p. 1322-1331
Abstract
Generative world models have demonstrated strong capability in modeling complex environment dynamics. However, existing approaches mainly rely on data pattern matching or physics-informed generation with externally given physical priors, failing to learn the intrinsic and coupled physical laws. Those challenges are particularly severe in fluid environments, where governing physical laws are implicit and jointly determined by multiple related physical factors. To solve these challenges, we propose the Physics Attention World Model (PAWM), a physics-learned generative world model that autonomously learns and decorrelated informs latent physics from raw fluid observations. Specifically, PAWM comprises two core modules: i) the Physics Attention module, which learns disentangled latent physics and achieves physics-based consistent constraints, and ii) the Decorrelated Physics-informed Prediction module, which first conducts decorrelation reweighting to quantify the contribution of different physical factors in video generation with theoretical support, then performs state transition and prediction conditioned on these learned and decorrelated latent physics. Extensive experiments on three fluid benchmarks demonstrate that PAWM significantly outperforms existing generative world models with respect to accurate and physically consistent prediction across both simulated and real-world fluid scenarios, as well as maintaining strong generalization across diverse fluid properties.Keywords
Funding Information
- National Key Research and Development Program of China (2022ZD0115903)
This publication has 16 references indexed in Scilit:
- Autonomous Embodied AI2026
- FluidNexus: 3D Fluid Reconstruction and Prediction from a Single Video2025
- DiViCo: Disentangled Visual Token Compression for Efficient Large Vision-Language ModelIEEE Transactions on Circuits and Systems for Video Technology, 2025
- ScenarioDiff: Text-to-video Generation with Dynamic Transformations of Scene ConditionsInternational Journal of Computer Vision, 2025
- Disentangled Representation LearningIEEE Transactions on Pattern Analysis and Machine Intelligence, 2024
- Physics-informed machine learningNature Reviews Physics, 2021
- Acceleration of global N2O emissions seen from two decades of atmospheric inversionNature Climate Change, 2019
- Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equationsJournal of Computational Physics, 2019
- The Unreasonable Effectiveness of Deep Features as a Perceptual Metric2018
- An Introduction to Fluid Dynamics2000