AI In a New AI Paper, CMU and Google Researchers Redefine Language Model Outputs: How Delaying Responses with Pause Tokens Boosts Performance on QA and Reasoning Tasks
