Open X-Embodiment
Cross-embodiment robot dataset assembled from more than one million trajectories across 22 embodiments.
Identity and state
- Public project and dataset documentation Public openness · reviewed 2026-09-03
Reviewed relationships
| State | Related entity | Relation | Scope | Boundary |
|---|---|---|---|---|
| Verified | ← Octo | trained on | published training mixture | The paper states 800k trajectories from Open X-Embodiment. |
| Verified | ← OpenVLA 7B | trained on | published training mixture | The paper states 970k real-robot episodes from Open X-Embodiment. |
Canonical evidence claims
Claims are projected from authored tables. A dash remains an explicit missing value; it is never filled from a neighboring product or name match.
| Metric | Value | Class |
|---|---|---|
| Claim boundary | abstract uses 21 institutions; body uses 34 labs / 60 datasets. Both stay attributed. Not an evaluation suite [benchmarks-r24] | robot-training-corpus |
| Embodiment in the source | 22 embodiments: single arms, bi-manual robots, and quadrupeds [benchmarks-r24] | robot-training-corpus |
| Engine / renderer | RLDS / OXE distribution [benchmarks-r25] | robot-training-corpus |
| First public date | arXiv 2023-10-13 [benchmarks-r24] | robot-training-corpus |
| Kind | multi-embodiment robot-learning dataset pool [benchmarks-r24] | robot-training-corpus |
| Language-conditioned | language annotations vary by constituent dataset [benchmarks-r24] | robot-training-corpus |
| Medium | real robot trajectories pooled from many datasets [benchmarks-r24] | robot-training-corpus |
| Paper | arXiv:2310.08864 [benchmarks-r24] | robot-training-corpus |
| Primary metric family | pretraining corpus; not an eval leaderboard [benchmarks-r24] | robot-training-corpus |
| Public code | yes — official project publishes the dataset [benchmarks-r25] | robot-training-corpus |
| Public leaderboard | — | robot-training-corpus |
| Published scale | 1M+ trajectories; 22 embodiments; 60 datasets; 34 labs; 527 skills; 160266 tasks (body). Abstract: 22 robots, 21 institutions, 527 skills, 160266 tasks [benchmarks-r24] | robot-training-corpus |
Source ledger
- Research Open X-Embodiment — “Open X-Embodiment: Robotic Learning Datasets and RT-X Models”. Type: research paper. ID: arXiv:2310.08864. Retrieved: 2026-09-03. reviewed 2026-09-03
- Research Open X-Embodiment — official project site. Type: research project page. Retrieved: 2026-09-03. reviewed 2026-09-03
- Research Kim et al. — OpenVLA: An Open-Source Vision-Language-Action Model. Type: research paper. ID: arXiv:2406.09246. Venue: CoRL 2025. Retrieved: 2026-08-29. reviewed 2026-08-29
- Research Octo Model Team — Octo: An Open-Source Generalist Robot Policy. Type: research paper. ID: arXiv:2405.12213. Retrieved: 2026-08-29. reviewed 2026-08-29
What is not inferred
No fuzzy entity matching, adjacent-column borrowing, universal compatibility, commercial maturity, or checkpoint-to-product identity. Absence of a relation means unreviewed or unsupported here—not proven incompatibility.