SpeedAug: Policy Acceleration via Tempo-Enriched Policy and RL Fine-Tuning

ArXi:2512.00062v2 Announce Type: replace-cross Robotic policy learning for complex real-world manipulation tasks has seen rapid recent progress, enabled in large part by the ability to collect nstrations through human operation. However, policies trained from such nstrations often execute tasks far slowly than the robot's physical capabilities, as nstration data is collected under practical constraints that favor conservative, success-oriented trajectories over execution speed.