AnalysisRoboticsOctober 26, 2026

HuGe pre-trains humanoid VLA models on egocentric human and robot data

Read original source →arxiv.org

HuGe pre-trains vision-language-action models on egocentric human and robot data, using pseudo-proprioception and visual adaptation to enable whole-body humanoid control. The paper has 17 authors and 11 GitHub stars.

1 source

More stories today

Open the live feed