WildWorld

WildWorld introduces a large-scale dataset and benchmark for dynamic world modeling in generative action role-playing games (ARPGs), capturing explicit game states and player actions. It is designed for researchers and developers working on AI-driven world simulation, offering a standardized testbed to evaluate models that predict state changes and generate coherent interactive environments. This project is interesting because it bridges the gap between static game data and truly dynamic, action-responsive world modeling, enabling more immersive and adaptive AI in games.