out-of-domain generalization

This describes a model's ability to perform well on data that comes from different distributions or domains than the training data. Enhancing out-of-domain generalization is a challenge in building robust AI systems.

7 papers