What are the characteristics of overfitting and underfitting in neural networks? Overfitting occurs when the model performs poorly on both training and unseen data, while underfitting occurs when the model performs well on the training data but poorly on unseen data.