How employing document retrieval methods can empower Large Language Models to generalize past their training data and solve novel problems.