Glossary · Methodology

Model Drift

A gradual decline in a deployed model's real-world performance over time as the data it encounters diverges from its training distribution, requiring monitoring and periodic retraining or re-evaluation.