EurekaZang123MIT
Kino-Fail
A counterfactual benchmark for recovery-relevant failure attribution in quadrupedal navigation, containing 11 physics and sensing interventions across 191 RTX/PBR scenes with 22,836 physical units.
Downloads0
Episodes22836
Why This Matters for Physical AI
Kino-Fail provides a systematic benchmark for understanding failure modes and recovery strategies in quadrupedal navigation, enabling better attribution of failures to physical versus sensing interventions for robust embodied AI.
Technical Profile
- Modalities
- rgbproprioception
- Robot Embodiments
- quadruped
- Action Space
- joint_positions
- Environment
- simulation
- Task Types
- navigationfailure_recovery
- Episodes
- 22836
- Data Format
- JSONL
- Annotation Types
- failure_labelsaction_labels
- License
- MIT
Access
Need custom rgb data?
Claru builds purpose-built datasets for simulation applications with dense human annotations and quality assurance.
Request a Sample Pack