Reproduce “DragonCrawl: A Generative, Intent-Based Framework for Scalable Mobile End-to-End Testing” 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
Sowjanya Puligadda, Mengdie Zhang, Ali Zamani, Dhruva Dixith Kurra, Eric Chen, Juan Marcano
DragonCrawl V2 tests mobile journeys by giving GPT-4o the goal, current screen, action history, and available actions, then checking outcomes with visual assertions. It selects each next action from context rather than replaying screen matches. In Uber’s production comparison, its iOS pass rate was 91.6% versus V1’s 80%; this does not establish causation or transfer beyond Uber’s two apps.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.