![[2602.15592] Uni-Flow: a unified autoregressive-diffusion model for complex multiscale flows](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.15592] Uni-Flow: a unified autoregressive-diffusion model for complex multiscale flows
Summary
Uni-Flow presents a novel autoregressive-diffusion model that effectively simulates complex multiscale flows, enhancing the accuracy and speed of fluid dynamics modeling.
Why It Matters
This research addresses significant challenges in modeling spatiotemporal flows across various fields, offering a solution that combines autoregressive and diffusion techniques. The ability to perform faster-than-real-time simulations has profound implications for scientific research and practical applications in fluid dynamics and cardiovascular modeling.
Key Takeaways
- Uni-Flow separates temporal evolution from spatial refinement for better modeling of complex flows.
- The model achieves stable long-horizon rollouts while reconstructing high-resolution physical fields.
- It enables faster-than-real-time inference in cardiovascular simulations, significantly reducing computation time.
Physics > Fluid Dynamics arXiv:2602.15592 (physics) [Submitted on 17 Feb 2026] Title:Uni-Flow: a unified autoregressive-diffusion model for complex multiscale flows Authors:Xiao Xue, Tianyue Yang, Mingyang Gao, Leyu Pan, Maida Wang, Kewei Zhu, Shuo Wang, Jiuling Li, Marco F.P. ten Eikelder, Peter V. Coveney View a PDF of the paper titled Uni-Flow: a unified autoregressive-diffusion model for complex multiscale flows, by Xiao Xue and 8 other authors View PDF HTML (experimental) Abstract:Spatiotemporal flows govern diverse phenomena across physics, biology, and engineering, yet modelling their multiscale dynamics remains a central challenge. Despite major advances in physics-informed machine learning, existing approaches struggle to simultaneously maintain long-term temporal evolution and resolve fine-scale structure across chaotic, turbulent, and physiological regimes. Here, we introduce Uni-Flow, a unified autoregressive-diffusion framework that explicitly separates temporal evolution from spatial refinement for modelling complex dynamical systems. The autoregressive component learns low-resolution latent dynamics that preserve large-scale structure and ensure stable long-horizon rollouts, while the diffusion component reconstructs high-resolution physical fields, recovering fine-scale features in a small number of denoising steps. We validate Uni-Flow across canonical benchmarks, including two-dimensional Kolmogorov flow, three-dimensional turbulent channel inflow generat...
