Guardado en:
Detalles Bibliográficos
Autores principales: Song, Chan Hee, Song, Yiwen, Goyal, Palash, Su, Yu, Riva, Oriana, Palangi, Hamid, Pfister, Tomas
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:https://arxiv.org/abs/2510.04673
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866914394429980672
author Song, Chan Hee
Song, Yiwen
Goyal, Palash
Su, Yu
Riva, Oriana
Palangi, Hamid
Pfister, Tomas
author_facet Song, Chan Hee
Song, Yiwen
Goyal, Palash
Su, Yu
Riva, Oriana
Palangi, Hamid
Pfister, Tomas
contents Computer-using agents (CUAs) must plan task workflows across diverse and evolving applications, yet progress is limited by the lack of large-scale, high-quality training data. Existing datasets are narrow, static, and costly to annotate, while synthetic data often yields oversimplified or misaligned behaviors. We present Watch & Learn (W&L), a framework that converts readily available Internet videos of human computer use into executable UI trajectories at scale. Instead of directly generating actions or relying on handcrafted heuristics, we cast trajectory annotation as an inverse dynamics problem that predicts user actions from consecutive screen states, which simplifies learning and generalizes across domains. Through a task-aware retrieval and labeling pipeline, W&L yields over 53K high-quality trajectories that enhance CUAs both as in-context exemplars and as supervised training data. On OSWorld, it consistently improves general-purpose and specialized CUAs, while on WindowsAgentArena it achieves state-of-the-art performance among 7B-scale models under the 15-step limit. These results show that web-scale human demonstration videos can serve as a practical and scalable foundation for advancing real-world CUAs.
format Preprint
id arxiv_https___arxiv_org_abs_2510_04673
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Watch and Learn: Learning to Use Computers from Online Videos
Song, Chan Hee
Song, Yiwen
Goyal, Palash
Su, Yu
Riva, Oriana
Palangi, Hamid
Pfister, Tomas
Artificial Intelligence
Computer Vision and Pattern Recognition
Computer-using agents (CUAs) must plan task workflows across diverse and evolving applications, yet progress is limited by the lack of large-scale, high-quality training data. Existing datasets are narrow, static, and costly to annotate, while synthetic data often yields oversimplified or misaligned behaviors. We present Watch & Learn (W&L), a framework that converts readily available Internet videos of human computer use into executable UI trajectories at scale. Instead of directly generating actions or relying on handcrafted heuristics, we cast trajectory annotation as an inverse dynamics problem that predicts user actions from consecutive screen states, which simplifies learning and generalizes across domains. Through a task-aware retrieval and labeling pipeline, W&L yields over 53K high-quality trajectories that enhance CUAs both as in-context exemplars and as supervised training data. On OSWorld, it consistently improves general-purpose and specialized CUAs, while on WindowsAgentArena it achieves state-of-the-art performance among 7B-scale models under the 15-step limit. These results show that web-scale human demonstration videos can serve as a practical and scalable foundation for advancing real-world CUAs.
title Watch and Learn: Learning to Use Computers from Online Videos
topic Artificial Intelligence
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2510.04673