Data practitioners must often estimate treatment effects to determine the causal effect of a variable on an outcome of interest. To improve upon the robustness and efficiency of supervised causal inference techniques used to estimate average treatment effects and quantile treatment effects, Chakrabortty et al. develop a set of semi-supervised estimators that leverage a smaller labeled dataset (that contains observations of an outcome and a set of possibly high dimensional covariates) and a much larger unlabeled datasets (where the response to treatment is not observed). They demonstrate the advantages of their approach on both simulated and real data (the Stanford HIV Drug Resistance Database).