Reproduce “MAPLE: A Multimodal Mobile Agent with Playbook-Learned Experience” 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

Jun Wang, Aurora Wei, Lucas Spangher, Ming Zhang, Eugene Ie, Yiannis Kantaros

MAPLE helps mobile agents complete multi-step tasks by separating temporary task facts from reusable rules about interface behavior. It predicts each action’s screen effect, compares that prediction with the observed transition, and stores rules in a Playbook. On MobileWorld’s 44 interactive tasks, it achieves 68.18% success versus 52.27% for adapted M3A; this supports the combined system, not any single component.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.