Overview
Google DeepMind’s family of vision-language-action models for robotics, with RT-2 and successor research projects.
Details
RT-2 and the RT-X family are Google DeepMind’s vision-language-action models for robotics, demonstrating that web-scale pretraining can transfer to robot control. RT-2 was a landmark research result showing how LLMs can be adapted for robotics. Google DeepMind continues advancing robotics AI through this and successor research.