lamshiu/gui-world-model-data

Dataset Analysis

GUI World Model Data Interaction data for training small action-conditioned world models of graphical user interfaces: a model sees a screen and an action (click, drag, double-click, key) and predicts the next screen. Every app and game here was driven automatically in headless Chromium (Playwright); each record is screen -> action -> next screen. Raw data only: screenshots, app states, actions. No model features are included (they are recomputed from the screenshots with a… See the full description on the dataset page: https://huggingface.co/datasets/lamshiu/gui-world-model-data.