A robust is insensitive to assumption violations, i.e., holds even when the Probability model is incorrect.