Uni-VLaT adds touch to VLA policies: back-tap walking goes from 0% to 85% success
stepjamUK · x · 2026-09-29
Uni-VLaT, from Tsinghua, Beihang, HKU and CUC researchers, adds touch to pre-trained VLA policies for humanoid loco-manipulation.
- Standout result is Back-Tap Walking: tap the robot's back and it walks; stop tapping and it stops — zero visual signal involved. Success goes from 0% without touch to 85% with it
- Other tasks show incremental gains: adding a prediction objective on raw tactile input lifts average success from 68% to 75% across five tasks
- The method generalizes across GR00T and π0.5 backbones
More from Embodied
- Imagine 'uv install Cloth Folding': The Pip-Install-Your-Labor Meme — k7agar · 2026-09-29
- DIY outdoor robot U-BOT now runs off a tool battery, but its 24V-rated buck converter fails — _Stocko_ · 2026-09-29
- Walden Robotics says good evals are gold: judging robot policies on autonomy, interventions and transfer — adnothing · 2026-09-29
- User pays 30% premium over Hertz just to rent a Tesla with FSD — a clear PMF signal — yunta_tsai · 2026-09-29
- Agility Robotics shows off dexterous manipulation of varied objects at IROS — lukas_m_ziegler · 2026-09-29
- First Microducks robotic ducks being assembled in Bordeaux for testing and demos — andimarafioti · 2026-09-29