Learning GUI Grounding with Spatial Reasoning from Visual Feedback