YOUR LANGUAGE MODEL IS A Q-FUNCTION

Summary from the TLDR AI newsletter

A fairly technical RL paper that shows the theoretical underpinning of reward models and base models in language.