Reproduce “Jev-Mobile: Jev as an Executor for Mobile GUI Agents” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Linghua Zhang
Jev-Mobile lets a vision-language model set short-term goals while a typed decision model selects executable Android actions from each fresh accessibility-tree observation. This delegates low-level choices without reusing stale coordinates. On AndroidWorld, it succeeded on 0.79 of tasks versus 0.84 for a step-wise VLM; among successful trajectories, time and model API cost were lower in a system-level comparison.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.