What it measures
Not an evaluation benchmark. It provides training data for generalist policies; the RT-1-X and RT-2-X models are trained on the pooled data mixture.
Open X-Embodiment: Robotic Learning Datasets and RT-X Models
Open X-Embodiment is a large, cross-institution collection of real robot manipulation data: over 1 million trajectories from 22 robot embodiments, contributed by 21 institutions. The project also released RT-X models trained on the pooled data. It is a training resource rather than an evaluation benchmark, and its embodiments are single arms, bimanual robots and quadrupeds.
Not an evaluation benchmark. It provides training data for generalist policies; the RT-1-X and RT-2-X models are trained on the pooled data mixture.
Contributing labs pooled their existing robot datasets into one collection. The project page lists the included datasets in a shared spreadsheet and links the code repository. The arXiv abstract reports 527 skills and 160,266 tasks across the collection; these are the paper's own counts, not a single standardized task list.
Pooled cross-embodiment data is an input for many generalist robot policies, including humanoid ones. The project page does not mention any humanoid embodiment, so the dataset's direct relevance to humanoids is low.
Scoring
This record defines no leaderboard metrics.
Hardware
No in-dataset robot embodiments are linked.
22 robot embodiments described on the project page as spanning single robot arms, bimanual robots and quadrupeds. No humanoid is named, and the full list of individual robots was not opened, so no roster robot is linked.
Recorded outcomes
No result files yet
Evidence
Related evaluations