metadata
license: mit
task_categories:
- video-text-to-text
Ego2Web: A Web Agent Benchmark Grounded in Egocentric Videos
Project Page | Paper | Code
Ego2Web is the first benchmark designed to bridge egocentric video perception and web agent execution. It pairs real-world first-person video recordings with web tasks that require visual understanding, web task planning, and interaction in an online environment for successful completion.
The benchmark connects real-world human activities with web-based tasks, enabling research at the intersection of embodied perception and web interaction across diverse domains such as e-commerce, media retrieval, and knowledge lookup.
Citation
If you find this work useful, please cite:
@article{yu2026ego2web,
title={Ego2Web: Benchmarking Web Agents with Egocentric Video Grounding},
author={Yu, Shoubin and Shu, Lei and Yang, Antoine and Fu, Yao and
Sunkara, Srinivas and Wang, Maria and Chen, Jindong and
Bansal, Mohit and Gong, Boqing},
journal={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
year={2026}
}