思考以回忆：推理如何解锁大语言模型中的参数化知识

Thinking to recall: How reasoning unlocks parametric knowledge in LLMs

来源：Google Research · research.google