A new preprint on implicit Chain-of-Thought is out! Transformers Provably Learn to Internalize Chain-of-Thought