What does the term "overfitting" refer to in machine learning?

a$)$ When the model performs well on unseen data

b$)$ When the model fails to capture the underlying pattern in the data

c$)$ When the model performs poorly on training data

d$)$ When the model memorizes noise in the training data