We propose a toolbox of orthogonal survival learners for HTE estimation from time-to-event data under censoring.

We show that two-stage CATE estimators can be suboptimal for decision-making and propose a novel method to improve them.