Search for a command to run...
ThinkTwice: Jointly Optimizing Large Language Models for Reasoning and Self-Refinement