Teaching large language models how to absorb new knowledge

With a new method developed at MIT, an LLM behaves more like a student, writing notes that it studies to memorize new information.