GPT-6 Astra for robotics: a curated collection of papers, demos and evals
NielsRogge · x · 2026-09-20
NielsRogge compiled a Papers with Code collection tracking how the community uses GPT-6 Astra to control robots, in two distinct ways:
- As a policy (VLA-style): directly outputting actions from camera observations
- As an agent: with a dedicated harness for planning, writing code, and calling tools or lower-level policies
Notable entries include GPT-6 Astra on robotic manipulation (Sep 4), GPT-6 Astra as an Embodied Policy (Sep 13), In-Context Robot Learning with VLM Agents (Sep 16), Agent as Policy for Robotic Manipulation (Sep 11), and robot puzzle demos like unlocking the claw and three-ring threading.
The collection shows frontier models entering embodied AI along both end-to-end VLA and hierarchical agent routes.
More from Embodied
- Open-source dual-arm robot solves tangram puzzles with 90 demos, ~60% success — DanGrover · 2026-11-12
- Robotics has three regional playbooks: China volume, Europe inspection, US pilots — ingliguori · 2026-11-12
- 3,500 Laser-Armed AI Submarine Robots Patrol Norwegian Salmon Farms — philfung · 2026-11-12
- Model Autonomously Designs a Surgical-Robot-Style Robotic Arm CAD After 75 Minutes — _Stocko_ · 2026-11-12
- Dev releases local world model turning any image into interactive video, runs on RTX 3060 — lucidml_lover · 2026-11-12
- RoboRisk benchmark shows foundation models pose high harm rates in robotics; LEASH harness cuts risk — UoLiverpool · 2026-11-12