Optimizing Chain-of-Thought Reasoners via Gradient Variance Minimization in Rejection Sampling and RL Paper • 2505.02391 • Published 2 days ago • 20