Reviewed relationships 1
Verified
OpenVLA 7B trained on Open X-Embodiment Dataset
Scope published training mixture
Boundary The paper states 970k real-robot episodes from Open X-Embodiment.
Canonical evidence claims 11
Claims are projected from authored tables. A dash remains an explicit missing value; it is never filled from a neighboring product or name match.
Robot foundation models — policy weights, not URDF assets Policy Model 11 rows · 2 declared gaps
| Metric | Value and evidence |
|---|---|
| Code license | project and training code published; see repository [M2] |
| Kind | open VLA on Llama 2 + SigLIP/DINOv2 [R2] |
| Paper | arXiv:2406.09246 [R2] |
| Parameter count | 7B [R2] |
| Public API | Declared gap |
| Public benchmark table | paper tables only [R2] |
| Public date | arXiv 2024-06-13; CoRL 2025 [R2] |
| Published latency / rate | Declared gap |
| Stated embodiments in the source | Open X-Embodiment mixture (many arms) [R2] |
| Stated training data | 970k real-robot episodes from Open X-Embodiment [R2] |
| Weights | HF openvla/openvla-7b; backbone is Llama 2 (Community License) [M2] [R2] |
Source ledger 2
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.