WebUIBench

Dataset for the paper: WebUIBench: A Comprehensive Benchmark for Evaluating Multimodal Large Language Models in WebUI-to-Code

🏠 Homepage | 📖 arXiv

Introduction

We introduce WebUIBench, a large-scale and comprehensive benchmark designed to evaluate the WebUI-to-Code capabilities of Multimodal Large Language Models (MLLMs). WebUIBench comprises over 21K question-answer pairs derived from more than 0.7K real-world websites, encompassing 9 distinct subtasks. We conducted extensive experiments on 7 state-of-the-art closed-source and 22 prominent open-source MLLMs. Our key findings highlight the models' deficiencies in webpage generation tasks across various dimensions, including cross-modality reasoning, element localization, and webpage layout generation.

Contact

• Zhiyu Lin: zyllin@bjtu.edu.cn
• Zhengda Zhou: zhengdazhou@smail.nju.edu.cn
• Zhiyuan Zhao: tuzixini@gmail.com

🚩Citation

If you find this work is helpful, please kindly cite as follows. Thanks !

@article{webuibench,
  title={WebUIBench: A Comprehensive Benchmark for Evaluating Multimodal Large Language Models in WebUI-to-Code},
  author={Zhiyu Lin, Zhengda Zhou, Zhiyuan Zhao, Tianrui Wan, Yilun Ma, Junyu Gao, XueLong Li},
  journal={arXiv preprint arXiv:xx},
  year={2025}
}