Equilibrium Mismatch
Equilibrium mismatch is a failure mode in machine learning potential models in which the equilibrium distribution learned by the model deviates significantly from the target physical distribution. Even when local prediction accuracy is high, the global distribution may still be seriously incorrect.
Timeline (last 90 days)
The study identifies two core mechanisms that lead to sampling failure: dynamical trapping and equilibrium mismatch.
Equilibrium mismatch refers to the learned steady-state distribution itself being far from the target distribution; under certain explicit assumptions, even when the per-atom error in a locally symmetry-preserving model is very small, the distribution discrepancy still approaches its maximum value as the system size increases.
All Facts (2)
The study identifies two core mechanisms that lead to sampling failure: dynamical trapping and equilibrium mismatch.
50%UnverifiedEquilibrium mismatch refers to the learned steady-state distribution itself being far from the target distribution; under certain explicit assumptions, even when the per-atom error in a locally symmetry-preserving model is very small, the distribution discrepancy still approaches its maximum value as the system size increases.
50%