The fidelity score calculates the exact proportion of source facts accurately preserved by dividing the count of survived claims by the total sum of survived, lost, and distorted claims:
\[\text{Fidelity Score} = \left( \frac{|\text{survived}|}{|\text{survived}| + |\text{lost}| + |\text{distortions}|} \right) \times 100\]
When no claims are detected, it defaults to 100.0%, and individual URL scores are averaged across the run to produce the aggregate fidelity score.
flowchart LR
subgraph Input["Extracted Claims"]
S["Survived (S)"]
L["Lost (L)"]
D["Distortions (D)"]
end
subgraph Calculation["Scoring Formula"]
Formula["Score = (S / (S + L + D)) * 100"]
end
subgraph Output["Fidelity Metrics"]
URLScore["Per-URL Fidelity Score"]
AggScore["Aggregate Run Score (Mean)"]
end
S --> Formula
L --> Formula
D --> Formula
Formula --> URLScore
URLScore --> AggScoreReferenced by