OmniGuide steers diffusion VLAs at test time with differentiable 3D guidance, no retraining
thoma_gu · x · 2026-11-09
Presented at #CoRL2026, OmniGuide turns any VLM, geometry foundation model, and human guidance into a differentiable 3D energy function that steers action sampling of a frozen flow/diffusion VLA.
It targets known weaknesses of VLAs like π0.5 and GR00T: collisions in cluttered scenes, complex semantics, and learning from human demos. The method requires no extra data, no retraining, and is composable. Poster Mon 11/9, 5–6:30pm, #32.
More from Embodied
- EgoMo3R: Stanford's NeurIPS paper jointly reconstructs 3D motion and scenes from egocentric video — jcniebles · 2026-11-09
- Agent Buddy: an ESP32 animated companion that shows your AI coding agents' live status — DanWahlin · 2026-11-09
- Origami Robotics to Unveil New Direct-Drive Robot Hands and Data Glove at CoRL Party — ybisk · 2026-11-09
- Physical Intelligence to present π0.7 at CoRL oral with emergent robot capabilities — lucy_x_shi · 2026-11-09
- Fully open MolmoAct2 action reasoning model wins Best Paper at CoRL TAG — jmin__cho · 2026-11-09
- Dyna Robotics' Dynasaur Folds Clothes Nonstop at CoRL, Shrugging Off Every Interruption — JasonMa2020 · 2026-11-09