UW and NVIDIA Propose SGS for Zero Real-World-Data Robot Training

UW and NVIDIA introduced Success-Guided Sampling (SGS), a simple modification to task samplers that, combined with large-scale sim2real RL, enables robots to perform dexterous assembly tasks like gear fitting with zero real-world data or demonstrations.

2026-10-12 ~ 2026-10-12 · 2 related posts

• Zero real-world data: Success-Guided Sampling unlocks zero-shot dexterous robot manipulation via sim2real RL — Scobleizer · 2026-10-12
• UW & NVIDIA's SGS lets mega-scale RL solve quadruped locomotion and contact-rich assembly, distilling to real hardware — zhaoran_wang · 2026-10-12