NVIDIA researcher: train robot policies with privileged sim info, improving Pi 0.5
wellecks · x · 2026-10-30
NVIDIA researcher wellecks shares his team's work on simulation-based robot training: a key advantage of simulation is leveraging privileged information about the simulator during training. The team develops three methods for improving models, applicable both to frontier models and to training a Pi 0.5 model from Physical Intelligence. He notes many open challenges remain, but is excited about this scalable paradigm, arguing simulations will improve as coding agents get better.
Related event: NVIDIA Uses Privileged Simulator Info to Improve Robot Policies(2 posts)→
More from Embodied
- UCLA's ChairNav teaches autonomous wheelchairs to learn each rider's preferences — zhoubolei · 2026-10-31
- Amazon Echo kills local voice storage: everything goes to the cloud from March 28 — alex_verem · 2026-10-31
- HP paid $116M for Humane, then bricked every $699 AI Pin 10 days later — alex_verem · 2026-10-31
- Moxie, the $800 AI robot friend for kids, went silent overnight after shutdown — alex_verem · 2026-10-31
- Subvocal's YC-backed glasses decode silent speech at 200+ wpm, built in 4 months — garrytan · 2026-10-31
- Nomagic warehouse robots go from bimanual to trimanual, pause to solve tricky picks — m_wulfmeier · 2026-10-31