What would an “RLHF moment” for robotics take?
A blog post shared by one of its authors asks what’s missing from reinforcement learning for frontier robotics models. A separate roundup describes its argument as a call for a standard post-training recipe, like the one LLMs got.
TLDR
A blog post shared by one of its authors asks what it would take for robotics to reach the level of today’s large language models and beyond. A separate post summarizes the argument as a need for a standard post-training recipe, alongside other recent writing on robot models, fine-tuning and data collection.
Combined views
24.6K
2 Sources, first seen 17h ago