Verified65% confidenceFactExact time
DeepMind developed the VLA (Vision-Language-Action model) system, which treats the robot's physical actions on equal footing with visual tokens and language tokens.
3
Sources
65%
Confidence
Long-term
Relevance
7/18/2026
First Seen
Sources
Related Claims
UnverifiedVision-Language-Action (VLA) models are the latest paradigm for embodied intelligence, with representative works including Google DeepMind's RT-2 and Physical Intelligence's π0.82% similarUnverifiedDeepMind layered a Gemini model on top of its VLA, allowing users to issue instructions directly in natural language so the robot can carry on a conversation while working.81% similarUnverifiedDeepMind employs two collaborating modules: the ER model (essentially a VLM), which is better at reasoning, handles coordination and planning, while the VLA model is responsible for executing physical actions.81% similarVerifiedGoogle DeepMind's RT-2 was the first to directly use a vision-language model to generate robot action commands.79% similarUnverifiedVersion 1.5 of the DeepMind robotics system introduced two key capabilities: an agent component and a thinking component.73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/552105API
curl https://kongchang.com/api/v1/knowledge/claims/552105MCP
get_claim(id=552105)