Wavelet Flow Matching for Multi-Scale Physics Emulation
About
Accurate emulation of multi-scale physical systems governed by PDEs demands models that remain stable over long autoregressive rollouts while preserving fine-scale structures. Deterministic emulators produce overly-smoothed predictions, while generative approaches better capture details but are costly. Latent-space generative models have emerged as a compromise but with the additional cost of separately pre-trained autoencoders. We propose Wavelet Flow Matching (WFM), a novel generative emulator that overcomes current trade-offs between cost and skill by performing optimal-transport directly in the multi-scale wavelet space. Rather than learning a latent compression, WFM leverages the hierarchical structure of a U-Net to jointly predict transport velocities of a prescribed wavelet representation. On three challenging systems of chaotic fluid dynamics, WFM achieves superior long-horizon stability, accuracy and spectral coherence compared to state-of-the-art models. Our results clearly position the wavelet space as an effective training-free representation for generative emulation of complex physical dynamics.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| PDE System Rollout Forecasting | Rayleigh-Bénard | VRMSE (Rollout 01:32)0.779 | 14 | |
| PDE System Rollout Forecasting | Active Matter | VRMSE (01:16)0.849 | 14 | |
| PDE System Rollout Forecasting | Turbulent Radiative Layer 2D | VRMSE (Rollouts 01-32)0.544 | 13 | |
| Autoregressive rollout | Turbulent Radiative Layer 2D | CRPS (Rollout 01:32)0.415 | 11 | |
| Autoregressive rollout | Rayleigh-Bénard | CRPS (Rollout 01:32)0.003 | 11 | |
| PDE Emulation | TRL t_cool=0.03 | CRPS1.031 | 11 | |
| PDE Emulation | TRL t_cool=0.06 | CRPS1.038 | 11 | |
| PDE Emulation | TRL t_cool=0.10 | CRPS0.808 | 11 | |
| PDE Emulation | Rayleigh-Bénard | CRPS (Ra=10^6)0.054 | 11 | |
| PDE Emulation | Turbulent Radiative Layer 2D | Spectral Coherence RMSE0.148 | 11 |