Use Tools and Metrics to Evaluate Retrieval Performance GENERATIVE-AI-ENGINEER-ASSOCIATE Practice Questions (Page 4)

Questions 16–18

1. 16foundation · easy

   Which retrieval evaluation metric is most appropriate when you want to assess how quickly the first relevant document appears in the ranked list of retrieved results?

   • ANormalized discounted cumulative gain (NDCG)
   • BPrecision
   • CMean reciprocal rank (MRR)
   • DRecall
   Reveal Solution Discussion Select an answer first
2. 17foundation · easy

   In a RAG pipeline, you compute recall@5 for your retrieval component and find it is 0.6. What does this value indicate?

   • A60% of all relevant documents appear in the top 5 retrieved results
   • BThe retrieval system returns 5 documents for 60% of queries
   • C60% of the retrieved documents are relevant to the query
   • D60% of the queries have at least one relevant document in the top 5
   Reveal Solution Discussion Select an answer first
3. 18application · medium

   A machine learning engineer is evaluating a RAG system and wants to compute retrieval metrics like recall@k, precision@k, and MRR on a test set of queries with ground-truth relevance labels. They are working in a Databricks notebook and prefer to use a widely adopted Python framework. Which approach should they take?

   • AUse the `sklearn.metrics` module's `recall_score` function
   • BUse the `nltk.metrics` module's `precision` and `recall` functions
   • CUse the `ragas` library's `RetrievalEvaluator`
   • DUse the `evaluate` function from the Hugging Face `datasets` library
   Reveal Solution Discussion Select an answer first