A dataset of gameplay trajectories aligning rendered video with player inputs, camera motion, game state, events, goals, rewards, and outcomes for training world models and agents.
Datoric describes this dataset as action-conditioned gameplay trajectories for training world models and agents. The trajectories align rendered video with keyboard, mouse, or controller inputs, camera motion, player and object state, events, goals, rewards, and outcomes. The stated environment coverage includes navigation, action, platforming, driving, puzzles, resource collection, object interaction, inventory, and exploration. The official specification says subset-specific fields and delivery scope are confirmed during review; the Hugging Face repository provides the technical specification and sample metadata, while production data is supplied under a commercial license.
Be the first Cosign
Add it to your Products I use list. You can also share why you use it.
