Gap Between the Lines

A model achieves 99% accuracy on its training set but only 68% accuracy on a held-out validation set, while a simpler model on the same data gets 80% train and 78% validation accuracy. What does the first model's gap most strongly indicate?

The model has high bias and needs more capacity to fit the data

Overfitting: the model has memorized noise/patterns specific to the training set that don't generalize

Underfitting: the model is too simple to capture the underlying pattern

The validation set is mislabeled, since training accuracy is the more trustworthy signal