The training process of large models is increasingly appearing in public view.
Earlier, Luo Fuli, head of Xiaomi MiMo, publicly disclosed the training progress of MiMo-V2.6 during its reinforcement learning stage, giving outsiders a look at some concrete details of large model training.
Now, the Physical AI field has made a similar attempt.
On October 21, Shengshu Technology's world model team officially launched a public livestream of the entire Motubrain 2.0 training process, planned to last half a month, showing the public the training process and research progress of the embodied model.
This livestream focuses on one question: how is Physical AI actually achieved?
Physical AI, or physical artificial intelligence, aims to enable AI to understand the physical environment and complete real tasks through robots. Compared with large language models, Physical AI also needs to solve problems such as action decision-making, task execution, and cross-scenario generalization.
The Motubrain series is a family of embodied models developed by Shengshu Technology, dedicated to building a "brain" with general capabilities for robots and exploring capability generalization across different tasks, scenarios, and robot bodies.
Motubrain 2.0 will soon be open-sourced. Before the official open-source release, the team chose to first make its entire training process public, showing the outside world this model's journey from training and testing to capability iteration.
This public training covers four parts:
1. Pre-training: observing the impact of training scale on model performance and zero-shot generalization capability.
2. Post-training: further training on specific tasks to improve task execution capability.
3. Agent task execution: exploring whether the model can transfer existing experience to other tasks.
4. RSI (Recursive Self-Improvement) exploration: studying whether the model can evaluate and improve its own actions based on execution feedback.
What makes this livestream special is the attempt to open the "black box" of embodied model training.
In the past, the outside world learned about progress in embodied intelligence mainly through robot demos, evaluation results, and papers. How the model was trained and how its capabilities changed at different stages were usually hard to observe directly.
Motubrain 2.0 chose to make public the entire process from training and testing to capability iteration, hoping to let the outside world see how the model gradually forms its capabilities, as well as the failures, trial and error, and adjustments during training.
For the embodied intelligence industry, this approach makes the model development process more transparent. Researchers and developers can not only see the final results, but also have the opportunity to understand the specific problems encountered during training and how the team continuously improved the model through experiments and feedback.
At the same time, the team also hopes to use this livestream to share research experience, problems, and reflections with technical peers, jointly discuss key issues such as large-scale training, cross-task generalization, and autonomous improvement, and provide more reference for exploring implementation paradigms of Physical AI.
Of course, the related research is still at an exploratory stage, and specific results need to be verified through follow-up experiments and evaluations. Compared with only showing what a robot can ultimately do, making public how the model is trained and how its capabilities form is equally of research value.
Livestream entrance: www.motubrain.com/motubrain2.0-live
Official website link: www.motubrain.com
This article was provided by Shengshu Technology and is republished by QbitAI with authorization; the views belong to the original author.