murobotics/tangram-house-panda-330ep, measured from the public release on 13 September 2026. frame level, every row read
Nobody opens a robot dataset and watches all of it. We read every frame of every episode and report what is in there.
The opening of episode 0, rendered rather than teleoperated. Every episode in the release carries the same single task string, solve the tangram puzzle to assemble the house. That is one skill done 330 times, which is a real thing to own, and it is not the same as 330 episodes of coverage.
episode 0 of murobotics/tangram-house-panda-330ep, cut from the release's own video, 24 seconds. Nothing is re-rendered or sped up.
murobotics/tangram-house-panda-330ep carries 330 episodes and 691,855 frames, and every one of them carries the same single task string. That teaches one skill well and cannot support a breadth claim, which matters when a buyer compares episode counts across vendors.
| property | value |
|---|---|
| episodes | 330 |
| frames | 691,855 |
| frame rate | 10 fps |
| robot | panda |
| action dimensions | 8 |
| cameras | 2 |
| distinct task strings | 1 |
| shortest episode | 1,931 frames |
| median episode | 2,075 frames |
| longest episode | 2,626 frames |
A trailing hold is the number of frames at the end during which the commanded action does not change by more than 1e-4 on any component. Seconds come from the frame rate the release declares.
| trailing hold at least | episodes | share |
|---|---|---|
| 1 frames , 0.1 seconds | 209 | 63.3% |
| 5 frames , 0.5 seconds | 113 | 34.2% |
| 10 frames , 1.0 seconds | 10 | 3.0% |
| 30 frames , 3.0 seconds | 4 | 1.2% |
| 60 frames , 6.0 seconds | 1 | 0.3% |
The release is a meta directory plus one or more parquet files. Read data/chunk-000/file-000.parquet, group rows by the episode ranges in meta/episodes, and compute the same columns. No credentials.
This page is what a published release will tell you on its own. The full grading removes the part that scores worst and retrains on what is left, then reports whether the smaller corpus scores higher, with rollout counts and intervals. The one result measured on physical hardware is Stera-10M, two arms on a four arm SO-ARM101 bench, 21.9% against 36.9% task success across 1,270 rollouts.