G4_01259.mp4 Apr 2026
: It showcases RT-X , a general-purpose robotics transformer trained on this diverse data, demonstrating that "cross-embodiment" training (learning from many different types of robots) significantly improves the performance of individual robots. Where to Access
: The paper introduces the Open-X Embodiment Dataset, which contains over 1 million real-world robot demonstrations across 22 different robot platforms.
The video file is part of the Open-X Embodiment dataset, which is the largest open-source robotics dataset to date. g4_01259.mp4
: The specific video file you mentioned is part of the data hosted on the Open-X Embodiment GitHub and distributed via Google Cloud Storage.
: The Open-X Embodiment Collaboration (a massive joint effort including Google DeepMind, Stanford, Berkeley, and over 20 other institutions). : It showcases RT-X , a general-purpose robotics
: The Project Page provides visualizations, the dataset browser, and code links.
: You can read the complete study on arXiv:2310.08864 . : The specific video file you mentioned is
The primary research paper associated with this file is: Key Details about the Paper