Couldn’t agree more. This is exactly what we’re working on at Roboanalytica. Trying to build a benchmark for actual robots, not just another LLM benchmark disguised as robotics or something that only works in sim. Would love to hear what you think is missing from current real-world evals, or if anyone wants to get involved!
Humanoid robots have gone viral.
But what can they actually do?
Introducing RoboAnalytica. We test complete robots: hardware, software, and control working together.
Starting with the Unitree G1. Watch our first findings below.