Models from the paper "Fine-Tuning with Divergent Chains of Thought Boosts Reasoning Through Self-Correction in Language Models"