Consistent Labeling Across Group Assignments: Variance Reduction in Conditional Average Treatment Effect Estimation

ICLR 2026 Conference Submission17876 Authors

Keywords: conditional average treatment effect, consistent labeling, group assignments, potential outcomes, variance reduction

TL;DR: We propose an enhancement applicable to any existing CATE estimation algorithms to reduce prediction variance and thus improve test errors with theoretical justification.

Abstract: Numerous algorithms have been developed for Conditional Average Treatment Effect (CATE) estimation. In this paper, we first highlight an overlooked issue in CATE estimation: many algorithms exhibit inconsistent learning behavior for the same instance across different group assignments. We introduce a metric to quantify and visualize this inconsistency. Next, we present a theoretical analysis showing that this inconsistency indeed contributes to higher test errors and cannot be resolved through conventional machine learning techniques. To address this problem, we propose a general method called **Consistent Labeling Across Group Assignments** (CLAGA), which eliminates the inconsistency and is applicable to any existing CATE estimation algorithm. Experiments on both synthetic and real-world datasets demonstrate significant performance improvements with CLAGA.

Primary Area: causal reasoning

Submission Number: 17876