BenchmarkMobileWorld-Real / MobileWorld-Safety
MobileWorld
An open-source mobile AI cross-application execution evaluation benchmark released by Alibaba's Qwen team on GitHub, covering real-device testing and safety evaluation.
Timeline (last 90 days)
Sep 23
Mobile-Use Agent's published scores are MobileWorld 82.1, MobileWorld-Real 92.2, and AndroidDaily 97.2, with an end-to-end task success rate of 90%.
Expired50%
Sep 23
The benchmark suite released by the Qwen team includes MobilePA-Bench (planning), MobileWorld (cross-application execution, open-sourced on GitHub), MobileWorld-Real (real-device performance), and MobileWorld-Safety (safety evaluation).
Unverified50%