🧪 Test?View on arXiv
DeskForge: Dense Supervision from Desktop Environments for Computer-Use Agents
Said Gurbuz, Author 2, Author 3, Author 4, Author 5
GUI groundingdataset generationfine-tuningcomputer vision
2610.02320
Builder Relevance
1h ago80%
Abstract
DeskForge introduces a controllable desktop environment for generating large-scale supervision data to improve computer-use agents' performance in complex desktop scenes.
Reality Card
Core Claim
DeskForge-1M corpus significantly enhances GUI grounding and long-horizon task completion for vision-language models, with notable accuracy improvements on external benchmarks.
Method / Result
Qwen3.5-4B model accuracy increased by 11.51 percentage points on ScreenSpot-Pro.
Limitations
The reliance on specific desktop environments may limit generalizability to all user interfaces.
Paper to code
Verified implementation resources so builders can test the paper’s claims instead of stopping at the abstract.
No verified implementation link has been attached yet. AIBuzzHub will keep this panel separate from unverified search results.