RoboArchitect turns frozen VLA policies into memory-equipped robot agents, lifting success rates
_Hao_Zhu · x · 2026-11-03
RoboArchitect: turning VLA policies into agents with their own control and memory
Shengguang Wu introduced RoboArchitect, which addresses the failure modes of VLA policies by converting them into robotic agents with their own control strategies and persistent memory.
- It automatically designs and continually improves each agent around a frozen policy, with no weight updates
- Average success rates rise substantially on unseen scenes and tasks
- A thread with more details accompanies the announcement
More from Embodied
- Neuralink user navigates hospital to radiology using only thought control — GregCook2011 · 2026-11-03
- Stanford's RoboArchitect lifts frozen robot policies from 9.7% to 43.3% success without touching weights — dair_ai · 2026-11-03
- Robotics team to unveil system at RoboColosseum in 48 hours, demos era over — DJiafei · 2026-11-03
- Inside Starkey Omega AI+ Hearing Aids: Quad DNN, Better Speech, New Health Features — BrandonSawalich · 2026-11-03
- Agentic real2sim: Claude Fable + GPT-6 Astra rebuild scenes in RoboSimStudio in under 2 hours — berkeley_ai · 2026-11-03
- Ken Goldberg praises RoboSimStudio: LLMs tune robot sim params, CoRL accept in 3 weeks — berkeley_ai · 2026-11-03