Abhishek Pillai

Current benchmarks miss critical insights into how well models understand GUI transitions, with DDB revealing that even state-of-the-art models struggle with context-aware control and action identification.