New study provides insights on the effectiveness of LLM RAG and fine-tuning for topics that are not included in the model's training data.