AI Paper Atlas

Research map / Learning theory and algorithms / Prediction and decision losses

Robust Sequential Experimental Design for A/B Testing

Qianglin Wen, Xiangkun Wu, Chengchun Shi, Ting Li, Niansheng Tang, Yingying Zhang, Hongtu Zhu

ICML 2026 · Learning theory and algorithms › Prediction and decision losses · approach: Distributionally robust optimization · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add orthogonalized worst-case bias term to the design objective (robust RSD vs non-robust NRD) (ablation)sequential A/B design via dynamic programming · Contextual bandit simulation, additive and interactive effectsempirical MSE of ATEImproved
use robust sequential DP design instead of random, switchback, Neyman biased-coin and Bayesian biased-coin designsA/B testing allocation · Contextual bandit simulation (N = 21 to 42 days)empirical MSE of ATEImproved
use RSD when the sample is smallA/B testing allocation · Contextual bandit simulation, small Nempirical MSE of ATEImproved
use hierarchical across-day DP and within-day RL design instead of Neyman-type or switchback designs under carryover effectsdynamic A/B testing with carryover · Synthetic dynamic setting (T = 6, 12; three nonlinearity levels)empirical MSE of ATEImproved
use RSD on data calibrated to ride-sharing logsA/B testing allocation · Real-data-based simulation (additive and interactive effects)empirical MSE of ATEImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy