PhoneWorld: Scaling Phone-Use Agent Environments

ArXi:2605.29486v1 Announce Type: cross A central bottleneck for use agents is that controllable, reproducible environments covering real mobile behavior are hard to build at scale. Existing mobile-agent benchmarks have made important progress on evaluation, but they do not by themselves provide a scalable way to construct many new use environments. We present PhoneWorld, a reusable pipeline that converts real GUI trajectories and screenshots into controllable use environments, executable tasks, automatic verifiers, and.