6). rStar-Math – a new approach proposes three core components to enhance math reasoning: 1) a code-augmented CoT data synthesis method involving MCTS to generate step-by-step verified reasoning trajectories which are used to train the policy SLM…

→ View original post on X — @dair_ai