LLMs on robots jump real-world success from 16.7% to 97.3%
tri_dao · x · 2026-07-29
Tri Dao says putting an LLM “brain” on robots delivered a 4× SOTA jump without extra training.
The quoted post reports a simple composition idea: LLMs think, robot policies move, so connect the two. The claimed results are stark:
- Real robot: success rate improves from 16.7% to 97.3%
- Simulation (LIBERO-PRO): 12.8% to 53.3%
He says he was surprised by how well the approach works and argues that agentic robots may arrive sooner than expected.
More from Embodied
- Nori Robotics says its L3 will make human dexterity cheaper — chris_j_paxton · 2026-07-29
- Unitree G1 appears to hop out of a car in a striking robot demo — chris_j_paxton · 2026-07-29
- Tesla claims one Robotaxi equals seven Model Ys in economic output — JOBhakdi · 2026-07-29
- Interactive rings may need haptic output, not just voice input, for AI use — plopesresearch · 2026-07-29
- Hackathon team uses OpenAI Codex to drive a robot arm from natural language — davidfromkansas · 2026-07-29
- Bagel Labs releases WorldDiT, a sub-billion-parameter robot control model — EXM7777 · 2026-07-29