Overfitting

When a model learns training-data noise rather than the underlying pattern, performing well on training data but poorly on new data.