Smarter Way For Large Language Models To Think About Hard Problems

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 approaches that give LLMs this capability set a fixed computational…