Peking University and Qiyuan Robotics introduce HIL-UMI, robot-free embodied post-training data collection
Overview
Peking University and Qiyuan Robotics, a team led by Dong Hao, introduce HIL-UMI, a method that collects error-correction data for embodied models without a robot arm.
A handheld gripper records human demonstrations, and the model is connected to the collection process so it flags segments where its predicted actions diverge from the human's, prompting targeted recording.
The Leiphone report says HIL-UMI reaches 5.63 times the data collection efficiency of real-robot HG-DAgger, and an average task progress of 94.8 versus 58.0 for standard SFT under equal frame budgets. These figures are the report's own and come from a single source.
Written by AI from the articles below · updated Oct 20, 2:12 AM ET
Check the sources:
Article timeline
The articles in this story. Times are ET.
- LeiphoneNewsPeking University and Qiyuan Robotics propose HIL-UMI for robot-free embodied post-training
AIPeking University and Qiyuan Robotics propose HIL-UMI, a method that collects error-correction data for embodied models without a robot arm. A handheld gripper records human demonstrations, and the model flags segments where its predicted actions diverge from the human's, prompting targeted recording. The article reports 5.63x higher data collection efficiency than real-robot HG-DAgger and average task progress of 94.8 versus 58.0 for standard SFT under equal frame budgets.
Heat Trend
Not enough continuous observations to show a trend yet.