Salvato in:
Dettagli Bibliografici
Autori principali: Gonzalez-Pumariega, Gonzalo, Tu, Vincent, Lee, Chih-Lun, Yang, Jiachen, Li, Ang, Wang, Xin Eric
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:https://arxiv.org/abs/2510.02250
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866908810776412160
author Gonzalez-Pumariega, Gonzalo
Tu, Vincent
Lee, Chih-Lun
Yang, Jiachen
Li, Ang
Wang, Xin Eric
author_facet Gonzalez-Pumariega, Gonzalo
Tu, Vincent
Lee, Chih-Lun
Yang, Jiachen
Li, Ang
Wang, Xin Eric
contents Computer-use agents (CUAs) hold promise for automating everyday digital tasks, but their performance on long-horizon, complex problems remains unreliable. Single-rollout execution is brittle, with small errors compounding over time and leading to high variance in outcomes. While prior work has attempted to scale within a single rollout, such approaches have yielded limited gains. Scaling over multiple rollouts offers a more promising alternative but doing so effectively is challenging due to the difficulty of evaluating and selecting among long-horizon agent behaviors. We introduce Behavior Judge (BJudge), which addresses this challenge by representing agent executions as behavior narratives and comparing candidate behaviors at this level, substantially improving robustness and success rates. Using multiple rollouts, BJudge establishes a new state of the art (SoTA) in OSWorld at 72.6%, significantly outperforming prior methods and surpassing human-level performance at 72.36%, with comprehensive ablations validating key design choices. We further demonstrate strong generalization results to different operating systems on WindowsAgentArena and AndroidWorld. Crucially, our results highlight the strong effectiveness of scaling CUAs, when you do it right: effective scaling requires structured trajectory understanding and selection, and BJudge provides a practical framework to achieve this.
format Preprint
id arxiv_https___arxiv_org_abs_2510_02250
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Scaling Agents for Computer Use
Gonzalez-Pumariega, Gonzalo
Tu, Vincent
Lee, Chih-Lun
Yang, Jiachen
Li, Ang
Wang, Xin Eric
Artificial Intelligence
Computation and Language
Computer Vision and Pattern Recognition
Machine Learning
Computer-use agents (CUAs) hold promise for automating everyday digital tasks, but their performance on long-horizon, complex problems remains unreliable. Single-rollout execution is brittle, with small errors compounding over time and leading to high variance in outcomes. While prior work has attempted to scale within a single rollout, such approaches have yielded limited gains. Scaling over multiple rollouts offers a more promising alternative but doing so effectively is challenging due to the difficulty of evaluating and selecting among long-horizon agent behaviors. We introduce Behavior Judge (BJudge), which addresses this challenge by representing agent executions as behavior narratives and comparing candidate behaviors at this level, substantially improving robustness and success rates. Using multiple rollouts, BJudge establishes a new state of the art (SoTA) in OSWorld at 72.6%, significantly outperforming prior methods and surpassing human-level performance at 72.36%, with comprehensive ablations validating key design choices. We further demonstrate strong generalization results to different operating systems on WindowsAgentArena and AndroidWorld. Crucially, our results highlight the strong effectiveness of scaling CUAs, when you do it right: effective scaling requires structured trajectory understanding and selection, and BJudge provides a practical framework to achieve this.
title Scaling Agents for Computer Use
topic Artificial Intelligence
Computation and Language
Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2510.02250