TongUI: Internet-Scale Trajectories from Multimodal Web Tutorials for Generalized GUI Agents
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866917123012427776 |
|---|---|
| author | Zhang, Bofei Shang, Zirui Gao, Zhi Zhang, Wang Xie, Rui Ma, Xiaojian Yuan, Tao Wu, Xinxiao Zhu, Song-Chun Li, Qing |
| author_facet | Zhang, Bofei Shang, Zirui Gao, Zhi Zhang, Wang Xie, Rui Ma, Xiaojian Yuan, Tao Wu, Xinxiao Zhu, Song-Chun Li, Qing |
| contents | Building Graphical User Interface (GUI) agents is a promising research direction, which simulates human interaction with computers or mobile phones to perform diverse GUI tasks. However, a major challenge in developing generalized GUI agents is the lack of sufficient trajectory data across various operating systems and applications, mainly due to the high cost of manual annotations. In this paper, we propose the TongUI framework that builds generalized GUI agents by learning from rich multimodal web tutorials. Concretely, we crawl and process online GUI tutorials (such as videos and articles) into GUI agent trajectory data, through which we produce the GUI-Net dataset containing 143K trajectory data across five operating systems and more than 200 applications. We develop the TongUI agent by fine-tuning Qwen2.5-VL-3B/7B models on GUI-Net, which show remarkable performance improvements on commonly used grounding and navigation benchmarks, outperforming baseline agents about 10\% on multiple benchmarks, showing the effectiveness of the GUI-Net dataset and underscoring the significance of our TongUI framework. We will fully open-source the code, the GUI-Net dataset, and the trained models soon. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2504_12679 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | TongUI: Internet-Scale Trajectories from Multimodal Web Tutorials for Generalized GUI Agents Zhang, Bofei Shang, Zirui Gao, Zhi Zhang, Wang Xie, Rui Ma, Xiaojian Yuan, Tao Wu, Xinxiao Zhu, Song-Chun Li, Qing Computer Vision and Pattern Recognition Building Graphical User Interface (GUI) agents is a promising research direction, which simulates human interaction with computers or mobile phones to perform diverse GUI tasks. However, a major challenge in developing generalized GUI agents is the lack of sufficient trajectory data across various operating systems and applications, mainly due to the high cost of manual annotations. In this paper, we propose the TongUI framework that builds generalized GUI agents by learning from rich multimodal web tutorials. Concretely, we crawl and process online GUI tutorials (such as videos and articles) into GUI agent trajectory data, through which we produce the GUI-Net dataset containing 143K trajectory data across five operating systems and more than 200 applications. We develop the TongUI agent by fine-tuning Qwen2.5-VL-3B/7B models on GUI-Net, which show remarkable performance improvements on commonly used grounding and navigation benchmarks, outperforming baseline agents about 10\% on multiple benchmarks, showing the effectiveness of the GUI-Net dataset and underscoring the significance of our TongUI framework. We will fully open-source the code, the GUI-Net dataset, and the trained models soon. |
| title | TongUI: Internet-Scale Trajectories from Multimodal Web Tutorials for Generalized GUI Agents |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2504.12679 |