According to a paper published on arxiv.org, researchers have introduced DragonCrawl, an AI-driven mobile testing system that uses large language models for continuous regression testing of mobile applications.
The system evolved from embedding-based similarity matching to generative intent-based reasoning using GPT-4o’s multimodal capabilities, according to the research. DragonCrawl achieves a 91.6% pass rate on iOS and 92.2% on Android across 1,013 automated tests running continuously in CI/CD pipelines, the paper reports.
According to arxiv.org, DragonCrawl addresses challenges in traditional End-to-End (E2E) testing frameworks, which “struggle with UI volatility, maintenance overhead, and cross-platform scalability.” Unlike prior LLM-based testing research focused on exploratory testing and crash detection, DragonCrawl “validates specific user flows on every code change, blocking commits that break critical functionality,” the paper states.
The system reduces test onboarding time from 96-120 hours to under 4 hours and has saved an estimated 27 developer years in test maintenance effort, according to the research. The paper describes the architectural evolution from V1 (semantic embedding matching) to V2 (generative intent-based reasoning) and discusses implementation challenges including token explosion and memory constraints.
According to arxiv.org, the system integrates multimodal vision for end-state detection and tool calling for backend state transitions, enabling comprehensive regression testing that bridges UI interactions with system state.