Which metric tells you whether an answer is supported by the retrieved context?
Why B is the answer
Groundedness
Groundedness measures support from source material. Fluency and coherence measure how well it reads.
In detail
- Groundedness measures whether an answer is supported by the retrieved context.
- A good RAG evaluation uses several metrics.
Use case
An illustrative scenario, not a real incident.
A bot gives a polished answer that its sources do not support. Groundedness measures that, where fluency scores it highly.
See it in different ways
flowchart LR A["Test dataset"] --> B["Map columns to evaluator inputs"] B --> C["Quality and safety evaluators"] C --> D["Compare with the baseline"] D --> E["Release decision"]
sequenceDiagram participant C as Change participant P as CI/CD pipeline participant R as Production participant T as Traces C->>P: Runs evaluators on a test dataset P->>R: Releases if it passes R->>T: Records real traffic T->>P: New failing cases
flowchart TD
Q{"What do you need to know?"}
Q -->|"Is it supported?"| A0["Groundedness"]
Q -->|"Does it read well?"| A1["Fluency"]
Q -->|"Does it hang together?"| A2["Coherence"]Key point: Groundedness compares answer and context.
Why the other options are wrong
- A. Fluency
- Fluency measures how well it reads.
- C. Coherence
- Coherence measures how well it hangs together.
- D. Throughput
- Throughput measures volume.