Search papers, labs, and topics across Lattice.
The paper introduces Uni-Flow, a unified autoregressive-diffusion model for spatiotemporal flows that decouples temporal evolution from spatial refinement. Uni-Flow uses an autoregressive component to learn low-resolution latent dynamics for stable long-horizon rollouts and a diffusion component to reconstruct high-resolution physical fields. Validated on benchmarks including Kolmogorov flow, turbulent channel inflow, and aortic coarctation simulations, Uni-Flow achieves faster-than-real-time inference of pulsatile hemodynamics.
Uni-Flow achieves faster-than-real-time modeling of complex multiscale flows by decoupling temporal evolution and spatial refinement, opening doors for deployable surrogates in scientific machine learning.
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 generation with a quantum-informed autoregressive prior, and patient-specific simulations of aortic coarctation derived from high-fidelity lattice Boltzmann hemodynamic solvers. In the cardiovascular setting, Uni-Flow enables task-level faster than real-time inference of pulsatile hemodynamics, reconstructing high-resolution pressure fields over physiologically relevant time horizons in seconds rather than hours. By transforming high-fidelity hemodynamic simulation from an offline, HPC-bound process into a deployable surrogate, Uni-Flow establishes a pathway to faster-than-real-time modelling of complex multiscale flows, with broad implications for scientific machine learning in flow physics.
