Testing frontier language models as robot controllers
A post highlights results on the trade-offs between thinking harder and acting slower when frontier language models control real robots.
TLDR
Calling frontier language models “quite strong robot policies,” a post describes an effort to move beyond scattered demos with a larger, systematic test on real robots. The author points to results exploring the trade-offs between thinking harder and acting slower.
Combined views
23.6K
6 Sources, first seen ago
