JensenYuan/DexTacWAM_cube_handover

Cube Handover

One of the six evaluation tasks in DexTacWAM. 100 episodes, 34,218 frames at 30 fps, on a Dexmate torso with two Sharpa Wave dexterous hands.

Pick up the cube with the right hand and hand it over to the left hand.

Passing a cube between the two hands. The releasing hand has to wait until the receiving hand has secured the object, which is signalled by contact.

Format

LeRobot v2.1. Parquet under data/chunk-000/, metadata under meta/. 29 GB.

feature	dtype	shape
head_img	image	(192, 256, 3)	PNG-encoded
left_wrist_img	image	(192, 256, 3)	PNG-encoded
right_wrist_img	image	(192, 256, 3)	PNG-encoded
tactile	uint8	(2, 5, 192, 256)	two hands x five fingers
state	float32	(90)
actions	float32	(150)
tactile_flow	float32	(2, 5, 24, 32, 4)	per-taxel flow, precomputed
deform	uint8	(2, 5, 192, 256)	deformation rendering

episode_provenance.json records, per output episode, which raw episode and frame range it came from and why it was split there.

Intended use

Stage 2 (world model) and stage 3 (action expert) training. The configs live under configs/cube_handover/ in the code repository, and the normalization statistics committed there are computed against exactly this data. Statistics from a different conversion will de-normalize actions incorrectly without raising an error.

The configs address this corpus by directory name, so unpack it as data/datasets_lerobot/20260806_cube_handover/ and keep the published episode ordering; the validation split is selected by episode index.

The released tactile encoder was trained on the separate 488 diverse episodes corpus, not on this data.

We do not release a trained action expert for this dataset. In our experiments, the stage 3 action expert is randomly initialized and trained from scratch using this data.

License

Apache 2.0, matching the DexTacWAM code. Parts of that repository are additionally CC BY-NC-SA 4.0 where they derive from Genie-Envisioner; that restriction applies to those source files, not to this data.

Citation

@article{dextacwam2026,
  title   = {DexTacWAM: A Visuo-Tactile World-Action Model for Dexterous Manipulation},
  author  = {Yuan, Haoran and Wang, Zekai and Shao, Boning and Lu, Haoran and
             Darrell, Trevor and Lourentzou, Ismini and Zhan, Wei},
  journal = {arXiv preprint arXiv:2609.24976},
  year    = {2026}
}