Meta researchers distill System 2 thinking into LLMs, improving performance on complex reasoning

A technique by researchers at Meta FAIR distills System 2 thinking into LLMs so that they can perform reasoning tasks with less resources