Which of the following statements best describes the relationship between overfitting and underfitting in the context of loss functions?

Overfitting happens when a model learns the noise in the training data, so its training loss drops while the validation loss rises, indicating it no longer generalizes. Underfitting occurs when the model is too simple, making both training and validation losses high because it cannot capture the underlying pattern. Thus, the loss curve for overfitting shows a large gap between training and validation loss, whereas underfitting shows a uniformly high loss on both sets. For example, fitting a straight line to a quadratic dataset will underfit, giving high loss everywhere, while a high‑degree polynomial will overfit, giving low training loss but high validation loss. The key idea is that overfitting reduces training loss at the cost of validation loss, while underfitting keeps both losses high.