GaP paper: graph-as-policy multi-agent harness makes robot automation reliable
rbhar90 · x · 2026-09-08
A Berkeley/Nvidia CoRL paper introduces Graph-as-Policy (GaP), a multi-agent coding harness that builds directed computation graphs (perception, planning, control) from the MORSL skill library and rehearses task graphs in internal simulation to close the reliability gap of model-free policies on Variational Automation tasks. Discussion: Goldberg notes harness design like GaP can further boost benchmarked models such as Astra; Letian Fu argues harnesses + tool calls (IK, SAM3) beat raw VLA on cost, speed and reliability, and that self-evolving harnesses and skill libraries will rapidly expand the capability Pareto frontier — robotics as the next frontier of multimodal agents.
More from Embodied
- Meta faces new lawsuit alleging 'perv glasses' footage was used to train AI without disclosure — jonerp · 2026-09-08
- $995 robot controlled by GPT-6 Astra zero-shots opening a cabinet door and self-corrects — DeryaTR_ · 2026-09-08
- GPT-6 Demo Shows End-to-End Real2Sim From Robot Footage in One Prompt — CyberRobooo · 2026-09-08
- One-shotting a Raspberry Pi UPS HAT CAD model with Claude in a few minutes — burhop · 2026-09-08
- TartanIMU: a light IMU foundation model for robotics, 36% ATE gain — rsasaki0109 · 2026-09-08
- Tesla Cybercab caught driving calmly in heavy rain, wipers stay relaxed — yunta_tsai · 2026-09-08