Grounded Action Model: building robot foundation models on 3D grounding instead of VLAs or video models
DJiafei · x · 2026-09-23
Ranjay Krishna and DJiafei introduce the Grounded Action Model (GAM), a new paradigm for robot foundation models built on top of a pretrained 3D grounding model — "ground first, then learn to act."
- The two existing routes both fall short: language-reasoning VLAs lack the precision robotic manipulation needs, while pixel-level video generation models (WAMs) are overkill for the task.
- GAM instead grounds actions in 3D space via a grounding model rather than relying on language models or video generators.
- Core claim: precision is vital for manipulation, and 3D grounding is a better foundation than language or video modeling.
Related event: GAM: Grounded Action Model Proposes 3D Grounding as Robot Foundation Base(6 posts)→
More from Embodied
- Open-source dual-arm robot solves tangram puzzles with 90 demos, ~60% success — DanGrover · 2026-11-12
- Robotics has three regional playbooks: China volume, Europe inspection, US pilots — ingliguori · 2026-11-12
- 3,500 Laser-Armed AI Submarine Robots Patrol Norwegian Salmon Farms — philfung · 2026-11-12
- Model Autonomously Designs a Surgical-Robot-Style Robotic Arm CAD After 75 Minutes — _Stocko_ · 2026-11-12
- Dev releases local world model turning any image into interactive video, runs on RTX 3060 — lucidml_lover · 2026-11-12
- RoboRisk benchmark shows foundation models pose high harm rates in robotics; LEASH harness cuts risk — UoLiverpool · 2026-11-12