Mind2Web_bbox_eval
收藏资源简介:
MultiUI是一个包含730万个样本的数据集,来自100万个网站,涵盖多种多模态任务和UI布局。该数据集在网页UI任务中表现出色,如在VisualWebBench上提高了48%,在Mind2Web数据集上的动作准确率提高了19.1%。此外,它还具有很好的泛化能力,适用于非网页UI任务和非UI领域,如文档理解、OCR和图表解释。
MultiUI is a dataset consisting of 7.3 million samples sourced from 1 million websites, covering diverse multimodal tasks and UI layouts. It demonstrates excellent performance in web UI-related tasks, achieving a 48% improvement on the VisualWebBench benchmark and a 19.1% increase in action accuracy on the Mind2Web dataset. Furthermore, it exhibits strong generalization capabilities and is applicable to non-web UI tasks and non-UI domains such as document understanding, OCR, and chart interpretation.
Mind2Web_bbox_eval 数据集概述
数据集简介
- 名称: Mind2Web_bbox_eval
- 来源: 用于论文《Harnessing Webpage UIs For Text Rich Visual Understanding》
- 主页: MultiUI Homepage
- GitHub: MultiUI GitHub
- arXiv: 论文链接
数据集内容
- 规模: 包含730万个样本,来自100万个网站
- 任务类型: 涵盖多模态任务和UI布局
- 应用:
- 在VisualWebBench上提升48%
- 在Mind2Web数据集上提升19.1%的动作准确率
- 泛化到非Web UI任务和非UI领域,如文档理解、OCR和图表解释
联系信息
- Junpeng Liu: jpliu@link.cuhk.edu.hk
- Xiang Yue: xyue2@andrew.cmu.edu
引用
@misc{liu2024harnessingwebpageuistextrich, title={Harnessing Webpage UIs for Text-Rich Visual Understanding}, author={Junpeng Liu and Tianyue Ou and Yifan Song and Yuxiao Qu and Wai Lam and Chenyan Xiong and Wenhu Chen and Graham Neubig and Xiang Yue}, year={2024}, eprint={2410.13824}, archivePrefix={arXiv}, primaryClass={cs.CV}, url={https://arxiv.org/abs/2410.13824}, }




