Gemini 4 Argon scores 87% on ERQA and controls robot arms

What the research found

DeepMind's robotics post reports Gemini 4 Argon scores 87% on ERQA and shows Spot and robot-arm control via MuJoCo practice and zero-shot control, with Real2Sim and hindsight labeling uses. Background for robotics teams; overlaps #768.

Claims and how well the evidence supports them

1. Argon scores 87% on the ERQA embodied reasoning benchmark 98%
2. Argon controls a Boston Dynamics Spot quadruped using MuJoCo digital twin practice 98%
3. Argon creates calibrated simulation twins from robot camera feeds, egocentric video or sketches 95%

Evidence

• deepmind.google https://deepmind.google/blog/gemini-4-argon-for-robotics

Actionability rated medium.