The authors extended MobileGym with 35 functional reimplementations of commercial-style apps, including persistent internal state and cross-app workflows.
Open training stack scales mobile agents across apps and tools
OpenMobile-2 combines controllable app simulations, verifiable training tasks and hybrid GUI-tool interaction in a shared mobile-agent framework.
Chinese Tech
Chenyang Yan · Yingying Zhang · Kanzhi Cheng · Qiushi Sun · Zhengyuan Pan · Zheng Ma · +18 more
Nanjing University · Xi’an Jiaotong University · National University of Singapore · SenseTime · East China Normal University
Research Digest··2 min read
Yan and colleagues built an open training stack for agents that operate smartphones, including a controllable environment spanning more than 110 apps, nearly 12,000 demonstration trajectories and over 2,000 reinforcement-learning tasks.
Why this paper
From SenseTime and 5 others
In one line
OpenMobile-2 builds versatile mobile agents using scalable simulated environments and hybrid GUI-tool actions.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ·No stated limitations found
- ✓Reports numbers on named benchmarks (3 benchmarks)
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§