A gap between the data a model was trained on and the data it now encounters, which grows the longer a model stays deployed unchanged.
Continue to AI University →