RAX From Harvard/NVIDIA/Berkeley Lifts Robot Success From 26% to 94% With Frozen Weights
jiqizhixin · x · 2026-11-06
RAX (preprint 'Scaling Online Exploration for Robot Agents', Harvard/NVIDIA/UC Berkeley/MIT/Stanford) tackles the retraining bottleneck: a coding agent reads observations, writes and revises perception/control routines, and keeps working notes plus a skill repository, while model weights stay frozen. Verified routines run directly; failures return control to the agent. On RoboDojo's plug-in-charger task, success on unseen layouts rises from 26.0% to 94.0% over exploration episodes.
More from Embodied
- Voice-to-app demo: PlaiPin device + Grok Gadget SDK builds apps without a keyboard — tetsuoai · 2026-11-06
- Modded MacBook Pro gets a working touchscreen — used to play Honor of Kings — xiaohu · 2026-11-06
- Roboticist shares Stanford-collab research directions and BARS talk — Fowe · 2026-11-06
- Open-Weights Robot-Use Harness: Qwen3.8 Scores 31.39, Beating GPT-6 Astra's 28.97 — aigclink · 2026-11-06
- Dev builds Salvo battleship game for dual-screen iPhone Duo with head-to-head and AI play — HankYeomans · 2026-11-06
- First-hand demo of Wayve AI Driver on San Francisco streets impresses rider — Sethwinterroth · 2026-11-06