Meet DigiData: Smarter Training and Testing for Phone‑Control AIs
Imagine telling your phone what you want—and an AI taps, types, and swipes to get it done. This paper introduces DigiData, a large, diverse, multi‑modal dataset designed to train general‑purpose mobile control agents. Instead of relying on unstructured user logs, the team systematically explores app features to craft goals, leading to richer tasks and longer, more realistic action sequences.
To track real progress, they also present DigiData‑Bench, a benchmark of complex, real‑world mobile tasks. The authors show that the popular “step accuracy” metric can be misleading for UI agents, and they propose dynamic evaluation protocols plus AI‑assisted scoring that judge whether the task was actually completed.
Why it matters: Better training data + better tests = faster, more reliable assistants that can navigate apps like we do—making everyday digital tasks smoother and more intuitive.
Paper: http://arxiv.org/abs/2511.07413v1
Paper: http://arxiv.org/abs/2511.07413v1
Register: https://www.AiFeta.com
AI MobileAgents Dataset Benchmark HCI UX MachineLearning Automation Smartphones
