This paper develops a principled and tractable framework for contextual decision-making with missing covariates that delivers stable and reliable out-of-sample performance.
May, 2026