Out-of-Distribution (OOD) Generalization

A research area in machine learning concerned with model performance when the training distribution and test distribution differ; conventional methods include domain generalization and invariant risk minimization.