Embodied Turing Machines: Code-Only Robot Policies

A new paper introduces Embodied Turing Machines and the Code-Only-As-Policy (COAP) paradigm, showing pure-code robot policies score 75 on RoboDojo without VLM/VLA, and explores recursive self-improvement in simulation.

2026-10-14 ~ 2026-10-14 · 3 related posts

• Embodied Turing Machine: exploring how far code-only policies (COAP) can go — liuziwei7 · 2026-10-14
• Code-only-as-policy beats VLA and agentic harness on RoboDojo, no VLM in the control loop — liuziwei7 · 2026-10-14
• Embodied Turing Machines paper explores stateful code for robot recursive self-improvement — _akhaliq · 2026-10-14