Tag
Geometric Action Model repurposes a geometric foundation model for robot policy learning, achieving 85.5% on LIBERO-Plus with 6.9 ms inference, 55× faster than baselines.