Robot Learning
Microsoft’s Rho makes the robot itself a stage of training
The September research release separates broad pretraining, embodiment adaptation, and task learning.
Subject desk / 3 analyses
Robot Learning
Demonstrations, shared data, and adaptable policies.
Microsoft’s Rho makes the robot itself a stage of training ↗From the desk
Robot Learning
Rho’s correction experiments put adaptation after deployment in focus
Microsoft’s latest manipulation research studies a small adaptive component while retaining a frozen action expert.
Robot Learning
Open robot models move the bottleneck from access to adaptation
Rho and pi0.5 show why released checkpoints are valuable foundations rather than complete industrial applications.