aiqbench analysis
This paper introduces Autoregression-Free Neural Operators (AFNO) to address error accumulation in long-horizon predictions for time-dependent PDEs by modeling continuous-time vector fields in a latent space via flow matching. By eliminating recursive autoregressive rollouts, the method demonstrates improved stability and reduced rollout errors across six different partial differential equations. This approach offers a significant architectural shift for scientific machine learning applications requiring extended temporal forecasting without compounding prediction drift.
Autoregression-Free Neural Operators for Time-Dependent PDEs
arXiv:2605.25413v3 Announce Type: replace Abstract: Neural operators learn mappings from function-dependent inputs to solutions, providing an effective framework for solving partial differential equations (PDEs). For time-dependent PDEs, existing methods typically perform long-horizon prediction through autoregressive rollout directly in high-dimensional physical field spaces, where each predicted state is recursively fed back as the input for the next step. Although effective for short-term prediction, this autoregressive rollout and the lack of continuous-time modeling lead to progressive error accumulation over long-horizon rollouts. In this work, we propose Autoregression-Free Neural Operators (AFNO), which map the time evolution of PDEs into a latent space and model continuous-time vector fields within it. AFNO uses flow matching to learn the latent vector field, thereby enabling continuous evolution over extended horizons, avoiding autoregressive rollout and capturing dynamics under varying parameter configurations through explicit conditioning on physical parameters. Theoretical analysis and extensive experiments on six PDEs demonstrate that AFNO improves long-horizon prediction stability and consistently reduces rollout errors compared with the baselines.
Abstract:Neural operators learn mappings from function-dependent inputs to solutions, providing an effective framework for solving partial differential equations (PDEs). For time-dependent PDEs, existing methods typically perform long-horizon prediction through autoregressive rollout directly in high-dimensional physical field spaces, where each predicted state is recursively fed back as the input for the next step. Although effective for short-term prediction, this autoregressive rollout and the lack of continuous-time modeling lead to progressive error accumulation over long-horizon rollouts. In this work, we propose Autoregression-Free Neural Operators (AFNO), which map the time evolution of PDEs into a latent space and model continuous-time vector fields within it. AFNO uses flow matching to learn the latent vector field, thereby enabling continuous evolution over extended horizons, avoiding autoregressive rollout and capturing dynamics under varying parameter configurations through explicit conditioning on physical parameters. Theoretical analysis and extensive experiments on six PDEs demonstrate that AFNO improves long-horizon prediction stability and consistently reduces rollout errors compared with the baselines.
Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Numerical Analysis (math.NA)
arXiv:2605.25413 [cs.LG]
(or arXiv:2605.25413v3 [cs.LG] for this version)
Full text under the publisher's open licence · 175 words
Related coverage
ANCHOR: Error-Controlled Adaptive Numerical Correction for Neural Operator Time Marching
arXiv · Jun 16, 2026
Green-Routed Neural Operators:\\Physics Determines Where the Network Reads
arXiv · Oct 7, 2026
Inertial Manifold Neural Operator for Dissipative Time-Dependent Partial Differential Equations
arXiv · Aug 25, 2026
Data-Efficient Time-Dependent PDE Surrogates: Graph Neural Simulators vs. Neural Operators
arXiv · Sep 30, 2026