While individual URLs vary widely based on extraction density and content complexity, aggregate fidelity scores benchmark the net reliability of a grounded answer by averaging per-URL metrics across all attributed sources.
flowchart TD
A[Grounded AI Output] --> B[Segment Attribution]
B --> C1[URL 1 Claims]
B --> C2[URL 2 Claims]
B --> C3[URL 3 Claims]
C1 --> D1[Fidelity Evaluator]
C2 --> D2[Fidelity Evaluator]
C3 --> D3[Fidelity Evaluator]
D1 --> E1[Score: 80%]
D2 --> E2[Score: 50%]
D3 --> E3[Score: 100%]
E1 --> F[Mean Aggregation]
E2 --> F
E3 --> F
F --> G[Aggregate Fidelity: 76.7%]