AI-900 Practice Question: Describe features of generative AI workloads on Azure

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

• ✗

  The model citing academic papers when asked about scientific topics

  Why it's wrong here

  Citing academic papers is one particular application of citation, but the RAG concept covers any retrieved source, including internal policies, support articles, or web pages. Scientific citation emphasizes scholarly reputation and intellectual credit, whereas RAG citation emphasizes provenance and fact-checking for whatever domain the system serves. Therefore, focusing only on academic papers mischaracterizes citation as a discipline-specific format rather than a general mechanism for answer verification.

• ✓

  Indicating which source documents support an answer — enabling verification and reducing hallucination risk

  Why this is correct

  Indicating which source documents support an answer is the core meaning of citation in a RAG system because it explicitly links each claim to the retrieved evidence that generated it. This lets users verify the response against the underlying documents, which both builds trust and reduces hallucination risk by forcing the answer to stay grounded in the retrieved context. It is not a post-hoc reference but an integral part of how the model constructs and articulates the answer.

• ✗

  Quoting user messages back to them to confirm the AI understood the question

  Why it's wrong here

  Quoting the user's own question back is a conversational clarification technique that helps confirm understanding, but it does not attribute the answer to any external evidence. Citation by contrast points to a third-party document or data source that supports a claim, making the basis of the answer inspectable. Restating user input is about dialogue management, not about grounding the response in retrieved information.

• ✗

  Copyright attribution when the model quotes text from its training data

  Why it's wrong here

  Copyright attribution concerns legal and ethical obligations when a model reproduces text from its training data, such as ensuring original creators are credited or licensing is respected. This is distinct from RAG citation, which attributes an answer to specific documents retrieved at inference time to ground the response. Training-data reproduction relates to the model's memorization, not to the dynamic sourcing that retrieval-augmented generation uses for verifiability.

Go deeper

Related to this question