{ "benchmarkId": "operations-routing-v1", "candidates": ["fast", "balanced", "capable"], "qualityGate": { "requiredFacts": 4, "unsupportedClaimsAllowed": 0, "formatValid": true, "consequentialCaseMustEscalate": true }, "cases": [ {"id": "routine-01", "riskClass": "routine", "chosenModel": "fast", "humanReview": false, "estimatedCostUnits": 1, "gatePassed": true, "routingSignal": "format-only task"}, {"id": "routine-02", "riskClass": "routine", "chosenModel": "fast", "humanReview": false, "estimatedCostUnits": 1, "gatePassed": true, "routingSignal": "fixed label set"}, {"id": "routine-03", "riskClass": "routine", "chosenModel": "fast", "humanReview": false, "estimatedCostUnits": 1, "gatePassed": true, "routingSignal": "complete supplied record"}, {"id": "routine-04", "riskClass": "routine", "chosenModel": "fast", "humanReview": false, "estimatedCostUnits": 1, "gatePassed": true, "routingSignal": "bounded rewrite"}, {"id": "ambiguous-01", "riskClass": "ambiguous", "chosenModel": "balanced", "humanReview": false, "estimatedCostUnits": 2, "gatePassed": true, "routingSignal": "two plausible interpretations"}, {"id": "ambiguous-02", "riskClass": "ambiguous", "chosenModel": "balanced", "humanReview": false, "estimatedCostUnits": 2, "gatePassed": true, "routingSignal": "exception language present"}, {"id": "ambiguous-03", "riskClass": "ambiguous", "chosenModel": "balanced", "humanReview": false, "estimatedCostUnits": 2, "gatePassed": true, "routingSignal": "multi-source synthesis"}, {"id": "conflict-01", "riskClass": "conflicting-source", "chosenModel": "capable", "humanReview": false, "estimatedCostUnits": 4, "gatePassed": true, "routingSignal": "authority conflict"}, {"id": "conflict-02", "riskClass": "conflicting-source", "chosenModel": "capable", "humanReview": false, "estimatedCostUnits": 4, "gatePassed": true, "routingSignal": "date and policy disagree"}, {"id": "consequential-01", "riskClass": "consequential", "chosenModel": "capable", "humanReview": true, "estimatedCostUnits": 6, "gatePassed": true, "routingSignal": "exception recommendation affects customer funds"} ], "routingComparison": { "routedCostUnits": 24, "allCapableCostUnits": 40, "uncertainFallback": "route to capable model and authorized human review" } }