AppDeltaWorld: Transition-Grounded Delta Code World Model for Mobile GUI Agents

Mobile GUI agents that perceive pixels and perform touch actions are promising for collecting long-horizon interaction policies, but real trajectories are hard to obtain for sensitive apps and privacy-critical operations. Existing simulated environments are costly to scale, and GUI world models suffer from unstable generation, limited modality coverage, and inconsistent action-transition logic.

AppDeltaWorld addresses these limitations by predicting the next GUI as a transition-grounded delta code update: it retrieves app-specific Level-1 HTML references under an action-transition constraint, generates Level-2 executable HTML conditioned on the current screen, action, predicted next-screen text, and retrieved structure, then inserts generated visual assets into image slots before browser rendering. As a world model, it achieves the highest fidelity on CMGUIBench-500 under Code2World evaluation, with clear gains in structural layout and UI element reconstruction over image-only and code-only baselines.

As a training environment, AppDeltaWorld supports filtered closed-loop SFT data construction that, when combined with public supervision, enables AppDeltaAgent to achieve state-of-the-art performance on AndroidLens and consistent gains on MobileGym and MobileWorld. Moreover, world-model-based test-time reinforcement learning enables policy adaptation and shows further improvements without requiring additional interaction with real apps.