OpenComputer: Verifiable Software Worlds for Computer-Use Agents
OpenComputer is a new framework designed to create verifiable software worlds for computer-use agents. It includes components such as app-specific state verifiers and a self-evolving verification layer to enhance reliability. The framework currently supports 33 desktop applications and aims to improve the automation of computer tasks.
- ▪OpenComputer integrates four main components for constructing verifiable software worlds.
- ▪The framework covers 33 desktop applications and 1,000 finalized tasks across various software categories.
- ▪Experiments indicate that OpenComputer's verifiers align more closely with human judgment compared to traditional evaluation methods.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Artificial Intelligence arXiv:2605.19769 (cs) [Submitted on 19 May 2026] Title:OpenComputer: Verifiable Software Worlds for Computer-Use Agents Authors:Jinbiao Wei, Qianran Ma, Yilun Zhao, Xiao Zhou, Kangqi Ni, Guo Gan, Arman Cohan View a PDF of the paper titled OpenComputer: Verifiable Software Worlds for Computer-Use Agents, by Jinbiao Wei and 6 other authors View PDF HTML (experimental) Abstract:We present OpenComputer, a verifier-grounded framework for constructing verifiable software worlds for computer-use agents.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv cs.AI.
Discussion
0 commentsMore from arXiv cs.AI
