You've trained a model that looks strong on the data used during development, but the team is worried it may not generalize well to new data. You need to explain how you would tell whether the model is memorizing patterns instead of learning signal.
How do you evaluate whether a model is overfitting?
Gap between training and validation or holdout performanceCross-validation performance stability across foldsProbability quality degradation, such as worse log lossSigns that model complexity is too high for the available signal