Data Drift
Definition
The decay of model accuracy over time as the distribution of production input data diverges from training data.
Example Case
A coding model struggling with APIs updated after its training cutoff date.
The decay of model accuracy over time as the distribution of production input data diverges from training data.