Do you really need proprioceptive info in robot learning? See our new paper!

VLA can understand tactile and surprisingly it can generalize, to a moderate level!

Can you collect less robot teleoperation data? Yes! And you can compliment that with human videos!

Our spatially anchored tactile VLA model can now do subtle manipulation now!

Having humanoid to do repetitive jobs is the dream that I have for a long time. Now we are making a steady step towards it!

Using a mocap suit to kick yourself in the balls with a robot is a great metaphor to close out 2025.

DeepSeek just fixed one of AI's oldest problems. (using a 60-year-old algorithm) Here's the story: When deep learning took off, researchers hit a wall. You can't just stack layers endlessly. Signals either explode or vanish. Training deep networks was nearly impossible.

me once FSD Unsupervised ships

very nice baseline!

This is very interesting! I do totally agree that problems should be recursively decomposed. Haven't read into the details, but context passing is a very important thing to deal with.

Congrats Jitendra! Hope to see what is coming next!