On September 10, Xiaomi Technology published an article announcing that the Xiaomi-Robotics-U0 Embodied World Model Training Framework and Model Weights have been fully open sourced to provide 4B and 34B scale Xiaomi-Robotics-U0 model weights, and also supports various tasks such as embodying scene generation, physical scene editing, single image generation, image editing, video generation, and graphic interlaced sequence generation, as well as supporting training and inference frameworks for research on world models, embodied intelligence, and robotics applications. We hope to lower the technical threshold and accelerate the exploration and implementation of world model technology in real robot scenarios through open models and codes.

On September 10, Xiaomi Technology published an article announcing that the Xiaomi-Robotics-U0 Embodied World Model Training Framework and Model Weights have been fully open sourced to provide 4B and 34B scale Xiaomi-Robotics-U0 model weights, and also supports various tasks such as embodying scene generation, physical scene editing, single image generation, image editing, video generation, and graphic interlaced sequence generation, as well as supporting training and inference frameworks for research on world models, embodied intelligence, and robotics applications. We hope to lower the technical threshold and accelerate the exploration and implementation of world model technology in real robot scenarios through open models and codes.