PeterLiZijia
NFWM World-Model Artifacts
HDF5 datasets for world-model training and evaluation across robotics tasks (PushT, Reacher, Cube, TwoRooms, LIBERO-90) and Atari games (Boxing, Breakout), used by NFWM, LeWM, and SMWM methods.
Downloads0
Episodes57,763
Why This Matters for Physical AI
Provides diverse, large-scale world-model training data across robotic manipulation and game-playing tasks essential for learning generalizable forward models and planning in physical domains.
Technical Profile
- Modalities
- rgb
- Robot Embodiments
- robotic_armhumanoid
- Action Space
- end_effector_delta
- Environment
- simulation
- Task Types
- manipulationpushinggraspingreachingobject_manipulationgame_playing
- Episodes
- 57,763
- Data Format
- HDF5
- Annotation Types
- action_labels
Access
Need custom rgb data?
Claru builds purpose-built datasets for simulation applications with dense human annotations and quality assurance.
Request a Sample Pack