SimGym: A Framework for A/B Test Simulation in E-Commerce with Traffic-Grounded VLM Agents
SimGym is a new framework designed to simulate A/B tests in e-commerce using vision-language model agents. It aims to streamline the testing process, reducing the time required for experiments from weeks to under an hour. The framework has shown strong alignment with real buyer behavior, enhancing the efficiency of e-commerce storefront evaluations.
- ▪SimGym utilizes a traffic-grounded persona generation pipeline to create buyer archetypes from clickstream data.
- ▪The framework includes a live-browser agent architecture that allows for coherent shopping sessions across different storefronts.
- ▪Empirical results indicate that SimGym agents achieve 77% directional alignment with observed shifts in buyer behavior.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Artificial Intelligence arXiv:2605.19219 (cs) [Submitted on 19 May 2026] Title:SimGym: A Framework for A/B Test Simulation in E-Commerce with Traffic-Grounded VLM Agents Authors:Han Li, Vibhor Malik, Zahra Zanjani Foumani, Alberto Castelo, Shuang Xie, Ailin Fan, Keat Yang Koay, Yuanzheng Zhu, Meysam Feghhi, Ronie Uliana, Zhaoyu Zhang, Angelo Ocana Martins, Mingyu Zhao, Francis Pelland, Jonathan Faerman, Nikolas LeBlanc, Aaron Glazer, Andrew McNamara, Zhong Wu, Lingyun Wang View a PDF of the paper titled SimGym: A Framework for A/B Test Simulation in E-Commerce with Traffic-Grounded VLM Agents, by Han Li and 19 other authors View PDF HTML (experimental) Abstract:A/B testing remains the gold standard for evaluating modifications to e-commerce storefronts, yet it diverts…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv cs.AI.
Discussion
0 commentsMore from arXiv cs.AI
