PI-JEPA: Label-Free Surrogate Pretraining for Coupled Multiphysics Simulation via Operator-Split Latent Prediction

PI-JEPA: Label-Free Surrogate Pretraining for Coupled Multiphysics Simulation via Operator-Split Latent Prediction explores A physics-informed AI framework that drastically reduces the need for expensive simulation data by pretraining on unlabeled physics parameters, enabling faster and more accurate multiphysics surrogate models.. Commercial viability score: 7/10 in Physics-Informed AI.