Agent as Policy: a general agent writes code to control real robots directly
Mengzhao Jia · hf · 2026-09-15
The Hugging Face project Agent as Policy for Robotic Manipulation shows a general-purpose agent acting directly as a robot control policy.
- The agent interprets visual input and writes executable programs to control a physical robot
- It revises actions based on physical feedback across diverse manipulation tasks
- Demonstrates the "LLM agent as robot policy" paradigm without task-specific training
More from Embodied
- AI passport hack: letting an AI model write the entire firmware for a dynamic card — oran_ge · 2026-09-15
- Steam Frame review knocks support cycle; author points to Apple's iPhone 11 getting iOS 27 — AndrewSchmidtFC · 2026-09-15
- Engineer uses AI image gen to create PCB silkscreen art, calls it a life hack — realwillreil · 2026-09-15
- Bemo Technology to list in Hong Kong as direct-drive modules surge 45x in two years — 创业邦 · 2026-09-15
- MinimalPhone2, a QWERTY 'focus phone' with a kill switch, crowdfunds $2.3M — 创业邦 · 2026-09-15
- Pure-Rust visloc-rs adds visual-inertial SLAM, runs 3.46x faster than COLMAP on CPU — rsasaki0109 · 2026-09-15