Reproduce “RelayAgent: Mobile Task Automation by Delegating Subtasks to In-App AI Assistants” 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
Jingsheng Yan, Fangnuo Wu, Mingkai Dong, Haibo Chen
RelayAgent delegates tasks to apps’ natural-language assistants, using an LLM to split requests into steps and a runtime-updated capability library to route supported work, with GUI control as fallback. Against a GUI baseline, it had higher success on three benchmarks, including 46% versus 31% on AndroidDaily. Without fallback, paired successful tasks averaged 1.8× faster completion and 8.8× fewer LLM tokens.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.