cua-lite preprocessed version of OpenBMB/CAGUI. Chinese-language Android mobile dataset contributing understanding (UI-element functional captioning + OCR, region-conditioned) and use (multi-step agent demonstrations across Chinese apps such as Eleme, WeChat, NetEase Music).
ds = load_dataset("cua-lite/CAGUI")… See the full description on the dataset page:
https://huggingface.co/datasets/cua-lite/CAGUI.