Search Results for "visual\"

...It aims to create an “AI phone agent” that can perceive on-screen content, reason about user goals, and execute sequences of taps, swipes, and text input via automated device control interfaces like ADB, enabling hands-off completion of multi-step tasks such as navigating apps, filling forms, and more. Unlike traditional automation scripts that depend on brittle heuristics, Open-AutoGLM uses pretrained large language and vision-language models to interpret visual context and natural language instructions, giving the agent robust adaptability across apps and interfaces.