CUA-Master scales computer-use agents via open-world recursive improvements
Original title: CUA-Master: Scaling Computer-use Agents through Open-World Recursive Improvements
CUA-Master-27B scores 80.2 on OSWorld Verified and 36.9 on OSWorld 2.0 in GUI-only and hybrid benchmarks.
CUA-Master employs a nested double-loop recursive framework. The outer loop incorporates open-world data and model weaknesses to guide iterations. The inner loop generates diverse verifiable tasks from video seeds via visual-semantic reconstruction. A feedback-guided retry mechanism enables reliable trajectory collection. Extensive experiments show competitive performance across multiple benchmarks.
- Nested double-loop recursive improvement framework
- Outer flywheel uses open-world corpus and evolving model weaknesses
- Inner flywheel autocatalytically propagates task primitives
- Visual-semantic dual-path trajectory reconstruction from videos
- Competitive scores: 80.2 OSWorld Verified, 36.9 OSWorld 2.0
From the original article
1 min read at huggingface.co
Despite recent advances in CUAs, their development remains constrained by the limited diversity, controllability, and reliability of interaction data. Moreover, conventional optimization pipelines lack iterative feedback, while existing recursive improvement approaches often rely on self-generated experiences, limiting ability expansion and yielding diminishing returns.
By Xinlei Yu, Gen Li, Ji Zhao · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.