VLA policies

Researchers unveiled Uni-VLaT, a framework that adapts vision-language-action policies with whole-body tactile sensing so humanoid robots can perform loco-manipulation — walking while handling objects. By combining touch feedback across the robot's body with vision-based policies, the system aims to make humanoids steadier and more capable in physical tasks. Details remain limited to the paper listing so far, with little public reaction or performance data yet available.