89 upvotes
New model adds physics to robot instruction following
What is PDRVLA: A Vision-Language-Action Model with Physical Dynamics Representations?
PDRVLA is a Vision-Language-Action system that learns physical dynamics so it can predict how objects will move while following language commands. By avoiding costly video generation, it keeps the control loop fast enough for real-time robot actions.
Who it is for
Robotics developers and researchers can use it to build more responsive robots that understand the physics of their environment.
How to try it
Open the paper on Hugging Face: https://huggingface.co/papers/2610.37381