From Forecasting to Feedback: Evaluating Learned Dynamics for Tokamak Control
Learned temporal models are moving beyond offline forecasting toward interactive environments for decision making, yet logged-trajectory accuracy need not predict closed-loop behavior under policies optimized in the learned model. This forecasting-control gap matters in tokamak magnetic control, where coupled plasma-ci...
G. Shi, Si-Qi Ding, Xuan-He Wang et al.
· 0 citations