What might indicate that a model is overfitting?

High accuracy on training data coupled with low accuracy on validation data is a strong indicator that a model is overfitting. When a model is overfitting, it learns the details and noise in the training data to an extent that it adversely impacts its performance on new, unseen data. Essentially, the model becomes too specialized to the training data and fails to generalize its learning.

This situation typically arises when the model is overly complex relative to the amount of training data available or when irrelevant features are included, leading it to capture patterns that do not hold in the broader population represented by the validation dataset. The result is that while the model performs exceptionally well on the training data—because it memorizes it—it struggles to perform adequately on validation or test data, indicating poor generalization.

The other scenarios do not suggest overfitting; consistent performance across datasets indicates a well-generalized model, high performance on unseen data reflects effective learning and adaptation, and training time that is low compared to validation time does not directly inform us about a model’s ability to generalize, as it pertains more to computational efficiency rather than learning efficacy.