Thermodynamics-Informed GNN for Zero-Shot Digital Twins

A new framework enables zero-shot digital twins by coupling visual perception with a physics-informed Graph Neural Network. The architecture enforces energy conservation and non-negative entropy production through a metriplectic thermodynamic formalism, allowing adaptation to new geometries without retraining.