BiCAT mirrors the right arm's signed joints so both arms share motion tokens

DJiafei · x · 2026-11-18

Part 3 of the BiCAT thread explains how two token streams share a motion language: both arms share one workspace and are symmetric about the body, like human hands. BiCAT flips the right arm's signed joints, so mirrored motions receive the same tokens.

This lets the model learn each mirrored motion pattern once instead of separately for each arm, reducing the cost of learning bimanual coordination.

Original post →

More from Embodied

Embodied channel →