Explore a groundbreaking approach to creating high-quality datasets for mobile user interface (UI) modeling in this 15-minute conference talk from CHI 2024. Discover how researchers leverage Large Language Models (LLMs) to automatically mine UI data from Android apps, mimicking human-like exploration. Learn about the innovative noise filtering techniques and human annotation process used to ensure dataset quality. Gain insights into the resulting MUD dataset, comprising 18,000 human-annotated UIs from 3,300 apps, and its potential applications in UI element detection and retrieval tasks. Understand the significance of this research in establishing a foundation for future studies on modern, high-quality user interfaces.