AI RESEARCH
Controlling Transient Amplification Improves Long-horizon Rollouts
arXiv CS.LG
•
ArXi:2605.08856v1 Announce Type: new Autoregressive neural simulators now match classical solvers on short-horizon prediction of physical systems, yet their accuracy degrades rapidly when rolled out over long horizons. In this work, we identify transient amplification of perturbations around rollout trajectories as a structural mechanism driving rollout error.