Search papers, labs, and topics across Lattice.
The paper introduces Scale-Autoregressive Modeling (SAR), a novel generative approach for efficiently sampling fluid flow distributions on unstructured meshes. SAR leverages a coarse-to-fine hierarchical generation process, predicting low-resolution fields and progressively refining them with higher-resolution details conditioned on coarser predictions. Experiments on unsteady-flow benchmarks demonstrate that SAR achieves lower distributional error and higher per-sample accuracy compared to diffusion models, while also outperforming a flow-matching Transolver in speed by 2-7x.
Sampling high-resolution fluid flow distributions just got 7x faster, thanks to a hierarchical coarse-to-fine generative model that concentrates computation where uncertainty is highest.
Analyzing unsteady fluid flows often requires access to the full distribution of possible temporal states, yet conventional PDE solvers are computationally prohibitive and learned time-stepping surrogates quickly accumulate error over long rollouts. Generative models avoid compounding error by sampling states independently, but diffusion and flow-matching methods, while accurate, are limited by the cost of many evaluations over the entire mesh. We introduce scale-autoregressive modeling (SAR) for sampling flows on unstructured meshes hierarchically from coarse to fine: it first generates a low-resolution field, then refines it by progressively sampling higher resolutions conditioned on coarser predictions. This coarse-to-fine factorization improves efficiency by concentrating computation at coarser scales, where uncertainty is greatest, while requiring fewer steps at finer scales. Across unsteady-flow benchmarks of varying complexity, SAR attains substantially lower distributional error and higher per-sample accuracy than state-of-the-art diffusion models based on multi-scale GNNs, while matching or surpassing a flow-matching Transolver (a linear-time transformer) yet running 2-7x faster than this depending on the task. Overall, SAR provides a practical tool for fast and accurate estimation of statistical flow quantities (e.g., turbulent kinetic energy and two-point correlations) in real-world settings.
