Nvidia's ARC Teaches Robot Foundation Models to Reason: 5x Success on Reasoning Tasks, 91.7% Real-Robot Rate
arankomatsuzaki · x · 2026-10-10
Nvidia presents ARC, a reasoning recipe for robot foundation models that labels existing demonstrations with action-grounded traces and fine-tunes pretrained policies — no new demonstrations, no foundation-scale training, no architecture change. π0.5 jumps from 28.0% to 45.3% on RoboLab-120, gains reach +50pp on RoboLab-Reasoning-50, and π0.5 + ARC hits 91.7% real-robot success (up from 9.5%). Code, model, dataset and benchmark to be released.
Related event: Nvidia Unveils ARC: Robot Reasoning Success Boosted 5x Without New Data(3 posts)→
More from Embodied
- Hand-built robotic skin: AI asks to plug it into its head, reads touch in a minute — repligate · 2026-10-10
- Physical attention bias cuts cable-simulation prediction error by 15%+ — Avihai Giuili · 2026-10-10
- Wayve's AI Driver hits Turin roads in a Maserati Grecale just 4 weeks after integration — alexgkendall · 2026-10-10
- IROS 2026 Best Paper: humanoid robot learns tennis rallies from 5 hours of human motion data — qinzytech · 2026-10-10
- Tesla Teases Robotaxi "Floodgates Are Opening" as Expansion Signals Build — julianibarz · 2026-10-10
- Adaptive Sampling + Pure RL Yields Robot Policies Beyond Prior Locomotion Limits — jeffclune · 2026-10-10