GroundCUA is a large and diverse dataset of real UI screenshots paired with structured annotations for building multimodal computer use agents. It covers 87 software platforms across productivity tools, browsers, creative tools, communication apps, development environments, and system utilities. GroundCUA is designed for research on GUI grounding, UI perception, and vision-language-action models that interact with computers.… See the full description on the dataset page:
https://huggingface.co/datasets/CUAResearch/GroundCUA-ICLR-anon.
