STaR: Self-Taught Reasoner | Paper Explained

[@galmacky]

Thanks for making this video! The RL analogy was a bit of distraction in the paper, but there seems to be some link as to how we reinforce the new trajectories with improved rationales only when the old rationales fail.