Out-of-Distribution State

In machine learning, a state or input that deviates from the training data distribution. In such states, the output quality of models typically degrades significantly.