SGNO achieves stable long-horizon PDE rollouts by organizing autoregressive steps as spectral evolution updates with a constrained diagonal generator and learned correction, delivering a median 74.8% reduction in GMean100 error across ten APEBench tasks.
Title resolution pending
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
years
2026 2representative citing papers
Neural surrogates systematically under-resolve high-frequency content in multiscale PDEs due to spectral bias and irreversible coarse-graining losses, with success confined to low-dimensional manifolds and weather prediction as a non-generalizable case.
citing papers explorer
-
SGNO: Spectral Generator Neural Operators for Stable Long Horizon PDE Rollouts
SGNO achieves stable long-horizon PDE rollouts by organizing autoregressive steps as spectral evolution updates with a constrained diagonal generator and learned correction, delivering a median 74.8% reduction in GMean100 error across ten APEBench tasks.
-
Predictivity and Utility of Neural Surrogates of Multiscale PDEs
Neural surrogates systematically under-resolve high-frequency content in multiscale PDEs due to spectral bias and irreversible coarse-graining losses, with success confined to low-dimensional manifolds and weather prediction as a non-generalizable case.