Skip to main content

evaluate_sensing_grounding

Function evaluate_sensing_grounding 

Source
pub fn evaluate_sensing_grounding(
    claim: SensingClaim,
    evidence: &[SensingEvidence],
    policy: GroundingPolicy,
    authenticator: &dyn SensingEvidenceAuthenticator,
) -> GroundingDecision
Expand description

Ground a claim against at most 64 explicitly resolved evidence records.

Only exact (subject,predicate,object,context,metadata) matches with equal physical quantity and unit count as support. The evaluator never accepts an unauthenticated exact match. Among authenticated exact matches, it chooses the lowest known uncertainty, then citation ID, for deterministic output. Unknown/high uncertainty or absent required calibration returns Review; no exact authenticated support returns Denied.