A quality assurance engineer at Datum Corporation is reviewing the batch evaluation results for a Copilot Studio agent's generative AI responses. The evaluation report includes metrics: Coherence = 4.2/5, Groundedness = 2.8/5, Relevance = 3.9/5. Which metric indicates that the agent is generating responses that are not sufficiently supported by the knowledge source content?

## Why B is Correct Groundedness measures the degree to which a generated response is supported by the retrieved source context (the knowledge source content). A low Groundedness score (2.8/5) means the model is generating content that goes beyond — or contradicts — the information in the retrieved knowledge chunks. This is the hallucination-adjacent metric: low groundedness = the agent is 'making things up' relative to its sources.

## Why the Distractors Are Wrong A (Coherence): Coherence (4.2/5) measures whether the response is logically structured, grammatically sound, and internally consistent. A high coherence score means the response reads well — it does not evaluate factual grounding to sources.

C (Relevance): Relevance measures whether the response addresses the user's actual question. A 3.9/5 relevance is moderate — the agent is mostly on-topic but could improve. This does not specifically indicate unsupported claims.

D (All three together): While a combination of low metrics can suggest broader quality issues, the question specifically asks which metric indicates 'not supported by knowledge source content' — that is precisely Groundedness.

## Exam Tip Evaluation metric mnemonics: Coherence = 'does it read well?' Groundedness = 'is it from the source?' Relevance = 'does it answer the question?' Fluency = 'is the language natural?' Know each metric's specific focus.