Fig.1

From Issue #7 · 2026-06-22

Geometric Action Model for Robot Policy Learning

Jisang Han, Seonghu Jeon, Jaewoo Jung, René Zurbrügg, Honggyu An, Tifanny Portela, Marco Hutter, Marc Pollefeys, Seungryong Kim, Sunghwan Hong

arXiv:2606.17046 · 117▲ · cs.RO, cs.CV, cs.LG

View on arXiv →

Premium readers get an interactive explainer for this paper — a figure you can poke at, not just read.

Log in to unlockSee a live demo →

Fig. 1Interactive explainer · premium

What it is

GAM (Geometric Action Model) is a robot manipulation policy that repurposes a pretrained geometric foundation model (a transformer that infers 3D depth/geometry from images) by splitting it at an intermediate layer, using the shallow layers as an observation encoder and inserting a causal transformer that predicts future latent tokens conditioned on language, proprioception, and action history. The remaining deep layers of the same backbone then decode those predicted tokens into both future depth maps and robot action chunks in a single forward pass.

Why it matters

For practitioners building manipulation policies, this trades the multi-step denoising of diffusion-based policies for one feed-forward pass, cutting inference latency and parameter count while improving robustness to camera viewpoint changes, which is a common failure mode when a deployed robot's camera gets bumped or repositioned.

Practical takeaway

If you are working with vision-language-action or world-action models and hitting latency or camera-fragility problems, watch for policies built directly on geometric foundation model backbones (like DA3 or VGGT) rather than 2D vision-language backbones, since the 3D priors help most under viewpoint perturbation.

Key result

On LIBERO-Plus (out-of-distribution perturbations of the LIBERO benchmark), GAM scored 85.5% overall vs. 84.6% for the next best baseline, with a 9.7 percentage point gain in the camera-perturbation setting, at 1.4B params and 6.9ms inference (about 55x faster than the 382ms diffusion-based Cosmos-Policy). Note the standard LIBERO scores are saturated near 97-98% across most methods, so the differentiation is almost entirely in the perturbed setting.

Subscribe

Get the next issue.

Free. One email a week. Unsubscribe any time: no account, no dark patterns.