Groundedness and Relevance Quiz

What you'll be asked

1. What is the primary goal of assessing 'groundedness' in a Generative AI system?
2. If a model provides a factually accurate answer that is completely unrelated to the user's specific query, which metric is suffering?
3. You notice your RAG system is frequently producing 'hallucinations' where it cites facts not found in the source documents. Which optimization strategy is most appropriate?
4. Which of the following scenarios describes a 'False Positive' in the context of groundedness evaluation?
5. When building an automated pipeline for evaluating groundedness, why is using an LLM-as-a-judge (e.g., using GPT-4 to grade GPT-3.5) considered superior to simple keyword matching?

Answers and explanations come after you submit.