What is double descent?

Question: What does double descent describe in modern machine learning? A) A rule that deeper models always overfit more B) A method for training decision trees C) A loss function used only for SVMs D) A phenomenon where test error can decrease again after the interpolation threshold as model capacity grows Correct: D Explanation: Classical bias-variance intuition suggests test error rises after overfitting begins. Double descent shows that in some modern settings, test error may peak near the interpolation threshold and then decrease again as model capacity increases further. Topic: advanced ML / generalization / double descent