Cambridge, MA – To make large language models (LLMs) more accurate when answering harder questions, researchers can let the model spend more time thinking about potential solutions. But common ...