Why A is the answer

Groundedness evaluation against the retrieved content

Groundedness checks whether the answer is supported by the source. A fluent answer can still be fabricated.

In detail

• Groundedness checks whether answers are supported by the retrieved content.
• Run it on a test set before release.

See it in different ways

flowchart TD
  A["Foundry evaluators"] --> B["General: coherence, fluency"]
  A --> C["RAG: groundedness, relevance"]
  A --> D["Safety: hate, violence, protected material"]
  A --> E["Agents: tool call accuracy, task completion"]

mindmap
  root((Evaluators))
    Quality
      Relevance
      Coherence
      Fluency
    RAG
      Groundedness
    Agents
      Tool call accuracy
    Safety
      Violence
      Hate and unfairness
      Protected materials

flowchart LR
  S0["Question"] --> S1["Retrieved content"] --> S2["Answer"] --> S3["Groundedness score"]

Why the other options are wrong

B. Latency percentile analysis across the retrieval and generation steps

Latency measures speed.

C. Token count analysis per request and per user

Token counts measure size.

D. Fluency scoring on a sample of production answers

A fluent answer can still be invented.

Related questions

• Question 24A RAG app returns relevant passages, but answers add unsupported claims. Which evaluator targets this?
• Question 42Which control checks whether an answer is supported by source material?
• Question 26You must test how well answers address the question that was asked. Which evaluator fits?
• Question 23Why use RAG rather than fine-tuning to answer from company policies that change often?

Learn more

• Our lesson: Generative AI and agents
• Foundry Agent Service
• RAG in Azure AI Search