A statistical technique that removes bias from estimators while maintaining asymptotic normality, enabling valid inference on parameters of interest.