Robot guide · 2026-10-15

Which SO-101 datasets are worth training on?

49 public SO-101 datasets pass Datum's junk gate. Ranked here by check results, then size. Junk is already filtered out.

DatasetEpisodesLicenseChecks
szk1ck/so101-ycb-pickplace20,000apache-2.05/5 pass
dongyoonkim/so101-pi05-base-dataset13,969apache-2.0,mit6/6 pass
gpudad/so101_pick_cube10,993apache-2.05/5 pass
gpudad/so101_pick_cube_chunked10,990apache-2.05/5 pass
XvKuoMing/so101_chess6,095apache-2.05/5 pass
szk1ck/so101-pickplace-sim-mujoco5,000apache-2.05/5 pass
Cache-SCA/Isaaclab-so101_11task_baseCaP_3300epi3,300apache-2.05/5 pass
armnet/armnetbench_v01_lerobot_so1012,499apache-2.05/5 pass
armnet/armnetbench_v01_lerobot_bimanual_so1011,219apache-2.05/5 pass
ill337/so101-bin-lift-6cam-cylroom540apache-2.05/5 pass
zwan1003/pickplace_skills_v3_2380apache-2.05/5 pass
jasmine314342/picknplace-bimanual-360360apache-2.06/6 pass
liorbenhorin-nv/so101_vials_dr_randomized_grasp_sim250_cosmos10_20261008260apache-2.05/5 pass
HyeonseokE/rq2_turn_off_lever_ours_repeat_100_10fps100apache-2.05/5 pass
DylanSh/so101_groot_dataset_visual_check100apache-2.06/6 pass
zhengod/vla-grab-dataset-merged88apache-2.06/6 pass
HyeonseokE/rq2_turn_off_lever_ours_80_10fps80apache-2.05/5 pass
HyeonseokE/rq2_turn_off_lever_random_80_10fps80apache-2.05/5 pass
sreetz-nv/so101_sim_manual_newton_2026100157apache-2.05/5 pass
qb1t/so101_teleop_cubes50apache-2.06/6 pass
MoAIBo/giver_sidehandover_bluebrush_sceneA_subscene2_sujith21apache-2.05/5 pass
MoAIBo/receiver_sidehandover_bluebrush_sceneA_subscene2_qi21apache-2.05/5 pass
MoAIBo/receiver_sidehandover_wirestripper_sceneA_subscene2_qi21apache-2.05/5 pass
MoAIBo/giver_sidehandover_wirestripper_sceneA_subscene2_sujith21apache-2.05/5 pass
MoAIBo/receiver_crossover_bluebrush_sceneB_subscene1_leo21apache-2.05/5 pass