Mage-VL-Cowork-4B Achieves State-of-the-Art on OSWorld-Verified
Original title: Mage-VL-Cowork: A Full Training Recipe to Learn Hybrid Computer Use From Evolved Harnesses
The 4B hybrid CUA is trained through a five-stage curriculum using skill-conditioned generation workflows that synthesize 293806 tasks.
Effective computer-use agents combine visual GUI interaction with code and terminal execution within a single policy. Mage-VL-Cowork is a 4B hybrid CUA trained from a general multimodally aligned checkpoint through an end-to-end progressive five-stage curriculum across midtraining, SFT, and RL. Mage-CUA-Forge replaces per-task agent construction with reusable skill-conditioned generation workflows and synthesizes 293806 desktop and web tasks across 6915 skills while reducing recurring per-task expenditure by 82.6%. The model reaches 66.0% on OSWorld-Verified.
- Four billion parameter hybrid CUA model
- Five-stage curriculum across midtraining SFT and RL
- 293806 tasks synthesized across 6915 skills
- 82.6% reduction in recurring per-task expenditure
- State-of-the-art at its parameter size
From the original article
1 min read at huggingface.co
Effective computer-use agents (CUAs) should seamlessly combine visual GUI interaction with code and terminal execution within a single policy. We present Mage-VL-Cowork, a 4B hybrid CUA trained from a general multimodally aligned checkpoint through an end-to-end, progressive five-stage curriculum across midtraining, SFT, and RL.
By Xiaoyi Zhang, Haoqing Wang, Xinjie Zhang · 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.