Xiaomi Technology's official Weibo account has just announced the official open-sourcing of Xiaomi's embodied foundation model, Xiaomi-Robotics-1.
Xiaomi-Robotics-1 was pretrained on more than 100,000 hours of UMI data and post-trained using more than 10,000 hours of cross-embodiment data. This open-source release covers the complete process from real-robot post-training to model deployment, and also provides evaluation code for the relevant benchmarks.
The project website, open-source repositories, and other links are as follows:
Project website: https://robotics.xiaomi.com/xiaomi-robotics-1.html
GitHub: https://github.com/XiaomiRobotics/Xiaomi-Robotics-1
Hugging Face: https://huggingface.co/collections/XiaomiRobotics/xiaomi-robotics-1
According to an earlier report by IT Home, in July this year, Xiaomi Technology announced the official release of Xiaomi's embodied foundation model, Xiaomi-Robotics-1. The model was pretrained on 100,000 hours of real-world operation data and combined with cross-embodiment post-training to enable the embodied foundation model to work out of the box.
Xiaomi-Robotics-1 uses a two-stage training paradigm consisting of pretraining and post-training. During pretraining, the model learns general action-generation capabilities. Given the current visual observation and a language description, the model must predict a sequence of actions that moves the scene from its current state toward the target state.
