Skip to main content
  1. Public practice questions
  2. AB-620 Practice Exams | AI Agent Builder Associate
  3. AB-620 Practice Exam 3

Question 14 — AB-620 Practice Exam 3

A topic calls an API that returns 20 potentially relevant records for a user's question. The results are placed into a Custom data Table for a generative answers node. The response must be grounded in the three records most relevant to the user's query. What should the topic do before supplying the Table to the generative answers node? (Select one answer)

  1. Return all 20 records and allow the node to rank them internally

  2. Sort the records alphabetically and send the complete result set

  3. Filter and rank the API results so the three strongest matches occupy the first three records

  4. Place the most relevant records anywhere in the Table because the node evaluates every row

Correct answers

  1. Filter and rank the API results so the three strongest matches occupy the first three records

Explanations

Correct:

Filter and rank the API results so the three strongest matches occupy the first three records

The generative answers node uses only the first three records supplied through Custom data. Selection is positional rather than relevance-based, so retrieval logic should filter and rank the results before the Table reaches the node.

Wrong:

Return all 20 records and allow the node to rank them internally

The node does not rank the complete Table and then select its strongest matches. Records after the first three do not contribute to response generation.

Sort the records alphabetically and send the complete result set

Alphabetical ordering does not ensure that the first three records are the most relevant to the user's question. The upstream retrieval process should prioritize relevance.

Place the most relevant records anywhere in the Table because the node evaluates every row

The node uses the first three records rather than evaluating every supplied record. Relevant records positioned after those entries are ignored for response generation.