Dataset Viewer
Auto-converted to Parquet Duplicate
episode_index
int64
0
10k
language_instruction
stringclasses
15 values
actions
stringlengths
10.8k
32.2k
length
int64
16
47
0
Hold the bottle with left hand and unscrew cap with right hand
[[-0.21320833265781403, 0.22360335290431976, -0.26169273257255554, -0.09256470948457718, -0.1787301003932953, -0.2348720133304596, 0.08546410501003265, -0.15988686680793762, 0.13026739656925201, 0.0029614544473588467, 0.04051540791988373, 0.025082357227802277, -0.16975444555282593, -0.21980302035808563, -0.158686891198...
16
1
Lift the tray with both hands
[[0.24398036301136017, 0.12966644763946533, -0.156850203871727, -0.1503967046737671, 0.03168635442852974, 0.2690751552581787, -0.12432592362165451, -0.13481701910495758, -0.11293594539165497, -0.16545583307743073, -0.14290349185466766, -0.23200446367263794, -0.09536980092525482, 0.15181605517864227, -0.0913680493831634...
22
2
Fold the towel: left hand holds one end, right hand folds
[[-0.20278503000736237, -0.11843377351760864, -0.08777263760566711, -0.00958507414907217, 0.19160598516464233, -0.14913226664066315, 0.08542884141206741, 0.036817554384469986, -0.12320662289857864, 0.12905766069889069, 0.09704405069351196, 0.03746086359024048, 0.022118330001831055, -0.09224513173103333, 0.0781598463654...
23
3
Open the jar: left hand grips, right hand twists
[[-0.012104388326406479, -0.1811899095773697, -0.07770180702209473, -0.12993639707565308, -0.04376490041613579, 0.046827901154756546, 0.07532382011413574, -0.016302336007356644, -0.2507249712944031, 0.06365158408880234, 0.025902699679136276, -0.08257412165403366, 0.04473493620753288, 0.017903367057442665, 0.15475904941...
25
4
Pour from bottle in left hand into cup held by right hand
[[-0.11591386049985886, 0.198264017701149, 0.08374499529600143, -0.12632951140403748, 0.05065358057618141, -0.18724913895130157, 0.0018824661383405328, -0.1855705827474594, -0.028256677091121674, 0.024464337155222893, 0.16226664185523987, -0.06503113359212875, -0.008636224083602428, -0.011410870589315891, -0.0583939924...
30
5
Thread the needle: left holds needle, right threads
[[-0.06255549192428589, -0.041126228868961334, -0.22927768528461456, -0.2395547330379486, -0.2570488154888153, 0.03869779780507088, -0.12992152571678162, 0.09954619407653809, 0.12402822077274323, 0.10168260335922241, -0.03940510377287865, -0.08052095770835876, -0.04023811221122742, -0.2655876874923706, -0.1542270332574...
45
6
Tear the paper: both hands pull apart
[[0.09693332016468048, 0.10677260905504227, 0.17644426226615906, 0.07075612992048264, 0.03436609357595444, -0.03548239544034004, -0.18441516160964966, 0.13231492042541504, -0.11892671883106232, 0.02095405012369156, -0.3057563006877899, -0.002190621569752693, 0.21232371032238007, -0.27030348777770996, -0.141064211726188...
37
7
Open the book with both hands
[[0.20131531357765198, -0.026144811883568764, -0.18165309727191925, -0.30123117566108704, 0.20701031386852264, -0.09239612519741058, -0.13152699172496796, 0.28825780749320984, 0.09202782064676285, -0.20842306315898895, 0.10443582385778427, -0.24282187223434448, -0.162960022687912, 0.14258326590061188, 0.028391135856509...
28
8
Carry the box using both arms
[[0.006742107216268778, 0.0583556592464447, 0.10414911806583405, -0.056375473737716675, 0.14325523376464844, 0.283221036195755, 0.08353708684444427, 0.08398330956697464, -0.20823785662651062, 0.0390576608479023, -0.10805068165063858, -0.3246228098869324, -0.15697081387043, 0.2812148928642273, -0.07452549040317535, 0.08...
21
9
Left hand stabilize the board, right hand hammers
[[-0.1310800462961197, 0.013522996567189693, 0.06999056786298752, -0.0846204087138176, 0.05242487043142319, -0.1343231499195099, 0.13095267117023468, 0.017364919185638428, 0.03614923357963562, -0.10722359269857407, 0.025310412049293518, 0.069063700735569, 0.08311659842729568, 0.20411838591098785, 0.037753887474536896, ...
20
10
Handoff object from left to right hand
[[-0.02688627690076828, 0.07273119688034058, 0.06139806658029556, -0.004236487206071615, 0.0420801006257534, 0.20198464393615723, -0.05878877639770508, -0.020254865288734436, -0.07816802710294724, 0.22340728342533112, -0.1952587366104126, 0.01741814613342285, 0.06445985287427902, -0.21573612093925476, 0.070058345794677...
20
End of preview. Expand in Data Studio

neon-bimanual-10k

Bimanual manipulation — 10K episodes of synthetic 32-DoF joint trajectories for Neon VLA training.

Description

Two-arm coordinated tasks: hold+unscrew, lift tray, fold towel

Each episode contains:

  • language_instruction: Natural language task description
  • actions: JSON array of joint position trajectories (T × 32)
  • length: Episode length (timesteps)

Usage

import pyarrow.parquet as pq
table = pq.read_table("data.parquet")
df = table.to_pandas()
print(f"Episodes: {len(df)}, Columns: {list(df.columns)}")

Part of Neon VLA

Training data for Neon — open-source Vision-Language-Action model for humanoid whole-body control.

Total Neon dataset collection: 160K episodes across 8 datasets (~1.27GB)

Downloads last month
26

Models trained or fine-tuned on cagataydev/neon-bimanual-10k