phi-monster2026apache-2.0
Deconfounded spatial set (LIBERO)
A dataset of 535 episodes across 10 scenes designed to address spatial grounding shortcuts in robotic manipulation, using identical bowls as targets with distinct spatial phrases to deconfound identity from spatial relationships.
Downloads38
Episodes535
Hours1.0159
Technical Profile
- Action Space
- 7-DoF: action
- Environment
- lab
- Task Types
- manipulationgrasping
- Episodes
- 535
- Total Hours
- 1.0159
- Data Format
- LeRobot
- License
- apache-2.0
Access
Need custom physical AI data?
Claru builds purpose-built datasets for lab applications with dense human annotations and quality assurance.
Request a Sample Pack