Elamur Research⌕
☰
home / papers / 2026 / InfiniteWeb: Scalable Web Environment Synthesis for GUI Agent Training

InfiniteWeb: Scalable Web Environment Synthesis for GUI Agent Training

Paper · 2026 · Ziyun Zhang, Zezhou Wang, Xiaoyi Zhang et al.

Summary

Reinforcement-learning training of GUI agents needs many interactive environments. Existing benchmarks such as WebArena or OSWorld are built by hand and cover only a limited number of applications. InfiniteWeb addresses this by automatically synthesizing complete multi-page websites together with tasks and programmatic evaluators.

The pipeline starts from a short website seed and a reference design image, both mined from Common Crawl. It first produces tasks and derives from them a shared specification of data models and interfaces, so that all pages remain mutually compatible. The backend is then built with task-centric test-driven development (TCTDD), iterating until task-relevant tests succeed. In parallel, the frontend is generated to follow visual traits extracted from the design image. The evaluators monitor task-related variables, which yields dense rewards.

The system outperforms commercial coding agents on WebGen-Bench. Agents trained in its environments improve on OSWorld, Online-Mind2Web and MobileWorld. This makes cheap, scalable environment generation a practical alternative to manual benchmark construction.

Key points

  • Downstream gains from training on generated sites:
    • +6.9% on OSWorld, +5.7% on Online-Mind2Web and +3.9% on MobileWorld.
    • Multi-application OSWorld tasks rose from 3.8% to 9.7%.
  • WebGen-Bench results:
    • The full system with a GPT-5 backbone scores 85.6.
    • With GPT-4.1 it scores 77.4, still above Claude-Code using GPT-5 (75.8).
    • Removing TCTDD costs 5.0 points (80.6, roughly Codex level).
  • Dense vs. binary reward: UI-TARS-1.5-7B was run on 4,000 tasks with 4 trajectories each. Dense reward gave 767 tasks usable for GRPO, versus 174 with binary reward (4.4×).
  • Difficulty and discriminability of the environments:
    • Agents score much lower than on OSWorld: UI-TARS-1.5-7B 7.4% vs 24.5%, Agent S2 14.1% vs 27.3%.
    • The gap between the two agents is wider (6.7 vs 2.8 points).
  • Generation cost and availability:
    • About $1.93 per website (≈0.36M input / 0.34M output tokens), with a median of about 20 minutes per site.
    • 600 sites were produced in about 125 minutes in parallel.
    • Code and websites are released at github.com/microsoft/FIVE-UI-Evol.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Fields, topics and concepts

📚 Shelves 3The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksInformation Systems · 1,437 works

Cite

Cite as Ziyun Zhang, Zezhou Wang, Xiaoyi Zhang, Zongyu Guo, Jiahao Li, Bin Li and Yan Lu. 2026. InfiniteWeb: Scalable Web Environment Synthesis for GUI Agent Training. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.acl-long.1313

@inproceedings{zhang2026infiniteweb,
  title = {{InfiniteWeb: Scalable Web Environment Synthesis for GUI Agent Training}},
  author = {Ziyun Zhang, Zezhou Wang, Xiaoyi Zhang, Zongyu Guo, Jiahao Li, Bin Li and Yan Lu},
  booktitle = {Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2026},
  doi = {10.18653/v1/2026.acl-long.1313},
  url = {https://doi.org/10.18653/v1/2026.acl-long.1313}
}

References

32 references, 11 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Saaket Agashe, Kyle Wong, Vincent Tu, Jiachen Yang, Ang Li, and Xin Eric Wang. 2025. Agent s2: A compositional generalist-specialist framework for computer use agents. Preprint, arXiv:2504.00906. — [arXiv:2504.00906](https://arxiv.org/abs/2504.00906) find
  2. Anthropic. 2025. Claude code. https://claude.ai/ code. ↗
  3. Bei Chen, Fengji Zhang, Anh Nguyen, Daoguang Zan, Zeqi Lin, Jian-Guang Lou, and Weizhu Chen. 2022. Codet: Code generation with generated tests. Preprint, arXiv:2207.10397. find
  4. Mark Chen, Jerry Tworek, Heewoo Jun, Qiming Yuan, Henrique Ponde de Oliveira Pinto, Jared Kaplan, Harri Edwards, Yuri Burda, Nicholas Joseph, Greg Brockman, and 1 others. 2021. Evaluating large language models trained on code. arXiv preprint arXiv:2107.03374.
  5. Karl Cobbe, Chris Hesse, Jacob Hilton, and John Schulman. 2020. Leveraging procedural generation to benchmark reinforcement learning. In International Conference on Machine Learning (ICML). find
  6. Yujia Li, David Choi, Junyoung Chung, and 1 others. 2022. Competition-level code generation with alphacode. Science, 378:1092–1097.
  7. Evan Zheran Liu, Kelvin Guu, Panupong Pasupat, Tianlin Shi, and Percy Liang. 2018. Reinforcement learning on web interfaces using workflow-guided exploration. In International Conference on Learning Representations (ICLR). find
  8. Xinyi Liu, Xiaoyi Zhang, Ziyun Zhang, and Yan Lu. 2025. Ui-e2i-synth: Advancing gui grounding with large-scale instruction synthesis. arXiv preprint arXiv:2504.11257. find
  9. Common Crawl Foundation. 2024. Common crawl. https://commoncrawl.org. ↗
  10. Zimu Lu, Yifan Yang, Haoran Ren, Haocheng Hou, Hangyu Xiao, Ke Wang, Wenji Shi, Aojun Zhou, Minghao Zhan, and Hongsheng Li. 2025. Webgen-bench: Evaluating llms on generating interactive and functional websites from scratch. arXiv preprint arXiv:2505.03733. — [arXiv:2505.03733](https://arxiv.org/abs/2505.03733) find
  11. Xiang Deng, Yu Gu, Boyuan Zheng, Shijie Chen, Sam Stevens, Boshi Wang, Huan Sun, and Yu Su. 2023. Mind2web: Towards a generalist agent for the web. Advances in Neural Information Processing Systems (NeurIPS).
  12. Yujia Qin, Yining Ye, Junjie Fang, Haoming Wang, Shihao Liang, Shizuo Tian, Junda Zhang, Jiahao Li, Yunxin Li, Shijue Huang, and 1 others. 2025. Ui-tars: Pioneering automated gui interaction with native agents. arXiv preprint arXiv:2501.12326.
  13. Wenyi Hong, Weihan Wang, Qingsong Lv, and 1 others. 2024. Cogagent: A visual language model for gui agents. In CVPR.
  14. Carlos E Jimenez, John Yang, Alexander Wettig, Shunyu Yao, Kexin Pei, Ofir Press, and Karthik Narasimhan. 2023. Swe-bench: Can language models resolve real-world github issues? arXiv preprint arXiv:2310.06770.
  15. Quyu Kong, Xu Zhang, Zhenyu Yang, Nolan Gao, Chen Liu, Panrong Tong, Chenglin Cai, Hanzhang Zhou, Jianan Zhang, Liangyu Chen, Zhidan Liu, Steven Hoi, and Yue Wang. 2025. Mobileworld: Benchmarking autonomous mobile agents in agent-user interactive, and mcp-augmented environments. Preprint, arXiv:2512… find
  16. Yaniv Leviathan, Dani Valevski, Matan Kalman, Danny Lumen, Eyal Segalis, Eyal Molad, Shlomi Pasternak, Vishnu Natchu, Valerie Nygaard, Srinivasan Venkatachary, James Manyika, and Yossi Matias. 2026. Generative ui: Llms are effective ui generators. Preprint, arXiv:2604.09577. — [arXiv:2604.09577](https://arxiv.org/abs/2604.09577) find
  17. Kaixin Li, Ziyang Meng, Hongzhan Lin, Ziyang Luo, Yuchen Tian, Jing Ma, Zhiyong Huang, and Tat-Seng Chua. 2025a. Screenspot-pro: Gui grounding for professional high-resolution computer use. In Proceedings of the 33rd ACM International Conference on Multimedia, pages 8778–8786. find
  18. Kuan Li, Zhongwang Zhang, Huifeng Yin, Rui Ye, Yida Zhao, Liwen Zhang, Litu Ou, Dingchu Zhang, Xixi Wu, Jialong Wu, and 1 others. 2025b. Websailor-v2: Bridging the chasm to proprietary agents via synthetic data and scalable reinforcement learning. arXiv preprint arXiv:2509.13305. — [arXiv:2509.13305](https://arxiv.org/abs/2509.13305) find
  19. Zhihong Shao, Peiyi Wang, Qihao Zhu, Runxin Xu, Junxiao Song, Xiao Bi, Haowei Zhang, Mingchuan Zhang, Y. K. Li, Y. Wu, and Daya Guo. 2024. Deepseekmath: Pushing the limits of mathematical reasoning in open language models. Preprint, arXiv:2402.03300.
  20. Chenglei Si, Yanzhe Zhang, Ryan Li, Zhengyuan Yang, Ruibo Liu, and Diyi Yang. 2025. Design2code: Benchmarking multimodal code generation for automated front-end engineering. Preprint, arXiv:2403.03163.
  21. StackBlitz Labs. 2024. Bolt.diy. Accessed: 2025-04-22. find
  22. Qiushi Sun, Kanzhi Cheng, Zichen Ding, Chuanyang Jin, Yian Wang, Fangzhi Xu, Zhenyu Wu, Chengyou Jia, Liheng Chen, Zhoumianze Liu, and 1 others. 2025. Os-genesis: Automating gui agent trajectory construction via reverse task synthesis. In Proceedings of the 63rd Annual Meeting of the Association for…
  23. Xuehui Wang, Zhenyu Wu, JingJing Xie, Zichen Ding, Bowen Yang, Zehao Li, Zhaoyang Liu, Qingyun Li, Xuan Dong, Zhe Chen, and 1 others. 2025. Mmbench-gui: Hierarchical multi-platform evaluation framework for gui agents. arXiv preprint arXiv:2507.19478. — [arXiv:2507.19478](https://arxiv.org/abs/2507.19478) find
  24. Laurie Williams, E. Michael Maximilien, and Mladen Vouk. 2003. Test-driven development as a defect-reduction practice. In 14th International Symposium on Software Reliability Engineering (ISSRE). find
  25. Jingxu Xie, Dylan Xu, Xuandong Zhao, and Dawn Song. 2025a. Agentsynth: Scalable task generation for generalist computer-use agents. arXiv preprint arXiv:2506.14205. find
  26. Tianbao Xie, Mengqi Yuan, Danyang Zhang, Xinzhuang Xiong, Zhennan Shen, Zilong Zhou, Xinyuan Wang, Yanxu Chen, Jiaqi Deng, Junda Chen, Bowen Wang, Haoyuan Wu, Jixuan Chen, Junli Wang, Dunjie Lu, Hao Hu, and Tao Yu. 2025b. Introducing osworld-verified. xlang.ai. find
  27. Tianbao Xie, Danyang Zhang, Jixuan Chen, Xiaochuan Li, Siheng Zhao, Ruisheng Cao, Toh Jing Hua, Zhoujun Cheng, Dongchan Shin, Fangyu Lei, and 1 others. 2024. Osworld: Benchmarking multimodal agents for open-ended tasks in real computer environments. arXiv preprint arXiv:2404.07972.
  28. Yiheng Xu, Dunjie Lu, Zhennan Shen, Junli Wang, Zekun Wang, Yuchen Mao, Caiming Xiong, and Tao Yu. 2024. Agenttrek: Agent trajectory synthesis via guiding replay with web tutorials. arXiv preprint arXiv:2412.09605. — [arXiv:2412.09605](https://arxiv.org/abs/2412.09605) find
  29. Tianci Xue, Weijian Qi, Tianneng Shi, Chan Hee Song, Boyu Gou, Dawn Song, Huan Sun, and Yu Su. 2025. An illusion of progress? assessing the current state of web agents. Preprint, arXiv:2504.01382. find
  30. Ziyun Zhang, Xinyi Liu, Xiaoyi Zhang, Jun Wang, Gang Chen, and Yan Lu. 2025. Ui-evol: Automatic knowledge evolving for computer use agents. Preprint, arXiv:2505.21964. — [arXiv:2505.21964](https://arxiv.org/abs/2505.21964) find
  31. Lianmin Zheng, Wei-Lin Chiang, Ying Sheng, Siyuan Zhuang, Zhanghao Wu, and 1 others. 2023. Judging llm-as-a-judge with mt-bench and chatbot arena. In NeurIPS.
  32. Shuyan Zhou, Frank F. Xu, Hao Zhu, Xuhui Zhou, Robert Lo, Abishek Sridhar, Xianyi Cheng, Yonatan Bisk, Daniel Fried, Uri Alon, and 1 others. 2024. Webarena: A realistic web environment for building and evaluating autonomous agents. In International Conference on Learning Representations (ICLR). find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/zhang-2026-infiniteweb-scalable-web-environment-synthesis/