SCOREAgents & automation
Humanoid Robot Teleoperation Data Evaluation
Anto Patrex@antopatrex1
This project evaluates humanoid robot teleoperation data using the JEV eval model to assess label consistency and annotation completeness, enabling better understanding of data trainability.
We ran @typesafeai 's JEV eval model on humanoid-robot teleop data from our deployment sites.
99% label consistency. 91% annotation completeness. Sub-action segmentation. Frame-accurate timing.
Robot data sells by the hour. Hours don't tell you if it's trainable. x.com/antopatrex1/status/21013731056672851