Post training for Robotics Foundation Models — Challenge Phase 1 Dataset This is the public Phase 1 dataset for the RSS 2026 Workshop & Challenge on Post training for Robotics Foundation Models . Workshop website: The dataset contains real robot bimanual manipulation trajectories on three benchmark tasks, collected on a bimanual YAM follower teleoperated by a GELLO leader arm. Every frame is timestamp aligned across joint state, action, and three synchronized camera streams at 60 Hz , so any modality can be sliced or resampled by a common time index. Dataset Layout Each leaf directory is a self contained LeRobot v2.1 dataset: Subset semantics: expert data — high quality human teleoperation demonstrations. failure data — trajectories where the baseline policy failed; useful as negative signal for post training. success and hil data — trajectories where the baseline policy completed the task, including episodes in which a human intervened mid rollout. The per frame mode label distinguishes autonomous vs. human controlled segments (see Human in the Loop Semantics). Per Task Duration All subsets are recorded at 60 Hz. Task Subset Episodes Frames Hours : : : insert mouse battery expert…
We use cookies for essential functionality and analytics. You can accept or reject analytics cookies.Cookie policy