Bimanual manipulation benchmark with 50 tasks and strong domain randomization; leaderboard policies train on 50 clean demos per task. Listing requires public code and weights, and top entries exceed 95% on the Easy setting.
Metric: Mean of Easy (clean) and Hard (domain-randomized) success, 100 trials per task. One result per model: RoboFYI re-run, then official leaderboard, then paper, then third party. Runs that deviate from the official protocol are listed on the model page but not ranked.