You are using an LLM-as-a-Judge to evaluate a RAG pipeline. What is the primary benefit of synthetically generating question-answer pairs, rather than relying s...