On March 10, Zhiyuan Robot officially released its first universal embossed base model - Genie Operator-1 (GO-1 for short). This innovative achievement has attracted widespread attention in the science and technology community, especially in the field of home service robots, bringing new possibilities to future housekeeping. According to the official introduction of Zhiyuan Robot, the GO-1 model uses deep learning of a large number of human videos to show its outstanding performance in performing multiple housework tasks, such as delivering water cups, making meals, and welcoming guests.
In terms of technical performance, GO-1 has significantly improved compared to the existing models, with an average success rate of 32%. After a series of tests, GO-1 has performed particularly well in specific tasks such as pouring water, cleaning the tabletop and replenishing beverages. Zhiyuan Robot said that this advancement is due to its newly proposed Vision-Language-Latent-Action (ViLLA) architecture. This architecture is able to make full use of high-quality datasets and large amounts of video data, thereby significantly enhancing the generalization capabilities of the model.

In addition, Zhiyuan Robot jointly announced at the end of last year that it had opened the million-level real machine dataset project AgiBot World with several institutions. This dataset covers more than 100 real scenarios, of which 40% are home scenarios and 20% are industrial scenarios. The release of this dataset provides important basic support for training and improving robot capabilities.
Zhiyuan Robot particularly emphasizes that the GO-1 model has several core features: first, it can learn from human videos, thereby improving the understanding of human behavior; second, it has powerful small sample rapid generalization ability, and can adapt to new tasks with very little data or even zero samples; finally, it can migrate between different robot forms to achieve the flexible application of "one brain, multi-shaped". Through continuous evolution, GO-1 can obtain data from actual execution, continuously optimize and improve its own capabilities.
The release of Zhiyuan Robot marks an important step towards the popularization of embodied intelligence by housework robots. According to Cinda Securities' analysis, the progress of AI technology is fundamentally promoting the development of technologies such as humanoid robots and autonomous driving, and the future prospects of humanoid robot industrialization are worth looking forward to.