AExGym: Benchmarks and Environments for Adaptive Experimentation
Not on the map. Why?
Abstract (arXiv)
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.
Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 30 of its 72 references that are in the map. First posted on arXiv 8 Aug 2024.
Builds on
30 of 72 references.
References · 30
- 2014On the Complexity of Best Arm Identification in Multi-Armed Bandit Modelscited by 280
- 2016Optimal Best Arm Identification with Fixed Confidencecited by 233
- 2020Differentiable Expected Hypervolume Improvement for Parallel Multi-Objective Bayesian Optimizationcited by 176
- 2018Deep Bayesian Bandits Showdown: An Empirical Comparison of Bayesian Deep Networks for Thompson Samplingcited by 166
- 2013lil' UCB : An Optimal Exploration Algorithm for Multi-Armed Banditscited by 148
- 2021Tianshou: a Highly Modularized Deep Reinforcement Learning Librarycited by 139
- 2016Simple Bayesian Algorithms for Best Arm Identificationcited by 116
- 2021Parallel Bayesian Optimization of Multiple Noisy Objectives with Expected Hypervolume Improvementcited by 115
- 2022KuaiRec: A Fully-observed Dataset and Insights for Evaluating Recommender Systemscited by 94
- 2016Tight (Lower) Bounds for the Fixed Budget Best Arm Identification Bandit Problemcited by 79
- 2020Open Bandit Dataset and Pipeline: Towards Realistic and Reproducible Off-Policy Evaluationcited by 77
- 2019Tuning Hyperparameters without Grad Students: Scalable and Robust Bayesian Optimisation with Dragonflycited by 66
18 more
- 2018RecoGym: A Reinforcement Learning Environment for the problem of Product Recommendation in Online Advertisingcited by 65
- 2018A Contextual Bandit Bake-offcited by 64
- 2020Inference for Batched Banditscited by 63
- 2023TorchRL: A data-driven decision-making library for PyTorchcited by 59
- 2017Improving the Expected Improvement Algorithmcited by 54
- 2020Efficient Nonmyopic Bayesian Optimization via One-Shot Multi-Step Treescited by 41
- 2015Thompson Sampling for Budgeted Multi-armed Banditscited by 23
- 2016Large-scale Validation of Counterfactual Learning Methods: A Test-Bedcited by 8
- 2020Policy Gradient Optimization of Thompson Sampling Policiescited by 8
- 2022Adaptive Experimentation in the Presence of Exogenous Nonstationary Variationcited by 7
- 2019Engineering for a Science-Centric Experimentation Platformcited by 6
- 2020Assessing External Validity Over Worst-case Subpopulationscited by 5
- 2021Adaptive Nonparametric Psychophysicscited by 5
- 2019Thompson Sampling with Information Relaxation Penaltiescited by 4
- 2024Best of Three Worlds: Adaptive Experimentation for Digital Marketing in Practicecited by 4
- 2021Datasets for Online Controlled Experimentscited by 3
- 2023Adaptive Experimentation at Scale: A Computational Framework for Flexible Batchescited by 2
- 2020Distilled Thompson Sampling: Practical and Efficient Thompson Sampling via Imitation Learningcited by 1
Built on by
Not cited yet.