Overfitting
When a model memorizes its training data and fails to generalize to new, unseen inputs.
An overfit model scores well on training data but poorly in the real world. It is countered with more and cleaner data, regularization (dropout, weight decay), early stopping, and held-out validation sets.