Robust Sequential Experimental Design for A/B Testing
Machine Learning
2026-05-14 v1 Machine Learning
Abstract
Experimental design has emerged as a powerful approach for improving the sample efficiency of A/B testing, yet existing designs rely critically on correctly specified models. We study robust sequential experimental design under model misspecification and develop a unified framework that covers both contextual bandit and dynamic settings. Theoretically, we prove that our design bounds the worst-case mean squared error of the estimated treatment effect. Empirically, we demonstrate the effectiveness of the proposed approach using synthetic and real-world datasets from a leading technology company.
Cite
@article{arxiv.2605.12899,
title = {Robust Sequential Experimental Design for A/B Testing},
author = {Qianglin Wen and Xiangkun Wu and Chengchun Shi and Ting Li and Niansheng Tang and Yingying Zhang and Hongtu Zhu},
journal= {arXiv preprint arXiv:2605.12899},
year = {2026}
}