New research uses A/B test data to guide adaptive experiment design

A new research paper explores using historical data from A/B tests to inform the design of adaptive experiments, particularly those employing contextual bandits. The study introduces a method that combines off-policy evaluation with controlled warm-start simulations to assess potential gains from adaptive policies. Findings suggest that adaptive, context-aware strategies can outperform fixed allocations when significant treatment effect heterogeneity exists, while offering minimal benefit otherwise. The research validates these conclusions on established open benchmarks, providing a practical framework for determining the value of adaptive experimentation. AI

RANK_REASON The item is a research paper submitted to arXiv detailing a new methodology for experimental design. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• A/B testing
• alphaXiv
• arXiv
• CatalyzeX Code Finder for Papers
• Contextual Bandits
• Criteo Uplift
• DagsHub
• Doubly-Robust Estimators of Treatment-Specific Survival Distributions in Observational Studies with Stratified Sampling
• Gotit.pub
• Hillstrom
• Hugging Face
• IArxiv Recommender
• Influence Flower
• João Victor Ferreira Alves Ferreira Alves
• Lalonde
• ScienceCast

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →