Yhn TechnologyRobotics first seen 9 h ago, last 17 min ago, peak #11
Frontier AI models score below 2/100 at humanoid control
Original: Frontier models score under 2/100 controlling a simulated humanoid
A new robotics benchmark, RoboBench from Vals AI, finds that today's frontier language models score under 2 out of 100 when tasked with controlling a simulated humanoid robot. The result underscores how far current AI remains from general physical control, despite strong performance on text and code tasks. Commenters are treating the low scores as a sobering reality check on claims that AI capabilities transfer easily to embodied, real-world-style settings.
Why now: The stark gap between frontier models' hype and their near-failure on embodied control tasks is prompting debate about AI's readiness for robotics.
Vals AIRoboBenchfrontier AI models
Rank over time, top of the chart is #1. 5 snapshots from 5 h ago to 17 min ago.
Evidence
- Frontier models score under 2/100 controlling a simulated humanoid · miyukimatthews · 7
API: https://socialmediatrends-api.osmike.com/v1/trends/2305561