AExGym: Benchmarks and Environments for Adaptive Experimentation

TL;DR

This paper introduces AExGym, a benchmark and open-source library for adaptive experimentation that addresses real-world challenges, aiming to improve practical robustness and facilitate methodological advancements in adaptive trial designs.

Contribution

It provides a realistic benchmark with datasets and challenges for adaptive experimentation, along with a modular library to support development and testing of adaptive algorithms.

Findings

Highlights practical challenges like non-stationarity and delayed feedback

Provides a benchmark to evaluate adaptive methods in real-world scenarios

Offers an extensible library for developing custom adaptive experimentation environments

Abstract

Innovations across science and industry are evaluated using randomized trials (a.k.a. A/B tests). While simple and robust, such static designs are inefficient or infeasible for testing many hypotheses. Adaptive designs can greatly improve statistical power in theory, but they have seen limited adoption due to their fragility in practice. We present a benchmark for adaptive experimentation based on real-world datasets, highlighting prominent practical challenges to operationalizing adaptivity: non-stationarity, batched/delayed feedback, multiple outcomes and objectives, and external validity. Our benchmark aims to spur methodological development that puts practical performance (e.g., robustness) as a central concern, rather than mathematical guarantees on contrived instances. We release an open source library, AExGym, which is designed with modularity and extensibility in mind to allow experimentation practitioners to develop custom environments and algorithms.