LLM scores 19/20 zero-shot on robot gripper task, beating specialized model's 1/20
willdepue · x · 2026-11-01
- willdepue reports a strong LLM controlling a robot arm scored 19/20 zero-shot on a gripper task, versus 1/20 for fable — with no sim training at all.
- Context: adamkarvonen argued LLMs doing basic pick-and-place is unsurprising (like a human with a joystick), and specialized robotics models still win on dexterity tasks like screwing in light bulbs.
- willdepue agrees dexterity is specialized-for-now, but predicts specialized models won't hold the lead long, given how strong astra's zero-shot performance already is.
More from Embodied
- "Doesn't need tactile" vs "won't benefit from tactile" are different worlds, researcher argues — _krishna_murthy · 2026-11-01
- Tesla FSD spots reckless street racer on blind corner, avoids crash with kids aboard — WilliamWangNLP · 2026-11-01
- TransfHAR at UIST 2026: self-supervised wrist motions enable on-watch custom activity recognition — realkaranahuja · 2026-11-01
- Ronie: free open-source AI agent that makes Ray-Ban Meta glasses actually smart — Odd-Sun6296 · 2026-11-01
- Robot team claims world's first skateboard kickflip, code and paper coming — Jan_R_Peters · 2026-11-01
- He built a working 3D-printed RC skid steer with Claude Code — 163 prompts, 493M tokens, 16 days — CheapWelder4303 · 2026-11-01