A steerable generalist robot foundation model in development that combines skills and generalizes across unseen tasks, robots, and environments.
Physical Intelligence introduced π0.7 in April 2026 to explore how robots can combine learned skills in new situations. The research focuses on a general-purpose policy that can follow unfamiliar language instructions and transfer experience between tasks. The team reported experiments involving new appliances and a robot folding laundry without task-specific folding data. These demonstrations examine compositional generalization: using pieces of prior experience together to solve a problem beyond the original training examples.
Be the first Cosign
Add it to your Products I use list. You can also share why you use it.
