What is overfitting in machine learning?

Example Answer

The way I'd explain it in an interview is this: Overfitting is when a model learns patterns that are too specific to the training data, including noise, quirks, or accidental correlations that do not hold on new data. As a result, the model can look excellent during training and still fail when it sees unseen examples.

A useful way to recognize it is the gap between training performance and validation performance. If the training score keeps improving while validation performance stalls or gets worse, the model is usually learning detail that does not generalize. That is why overfitting is less about memorization in a narrow sense and more about poor generalization.

What matters in an interview is not only knowing the definition, but being able to connect it back to how it changes modeling, evaluation, or deployment decisions in practice.

Common Poor Answer

A weak answer says overfitting means the model memorizes the training set, without connecting that to the practical symptom: poor performance on unseen data.