What is the main statistical role of cross-fitting?

Question: In double machine learning, nuisance functions are trained on K−1 folds and evaluated through an orthogonal score on the remaining fold. What is the main purpose of this cross-fitting procedure? A) To guarantee that every nuisance learner is unbiased on each fold B) To ensure that all observations receive identical nuisance predictions C) To select the nuisance learner with the minimum cross-validation error D) To separate nuisance fitting from score evaluation, reducing overfitting bias and restrictive empirical-process requirements Correct: D Explanation: Cross-fitting ensures that the score for an observation is evaluated using nuisance estimates trained without that observation. This weakens the dependence between nuisance estimation errors and the estimating equation. It is not primarily a model-selection procedure and does not make the nuisance predictions unbiased. Topic: advanced ML / cross-fitting / semiparametric inference