Verification Cascade
The Verification Cascade is a 4-stage confidence-gated system that ensures Aucert's false positive rate stays below 5%. Each stage increases verification rigor (and cost) — most results resolve at Stage 1.
Stage 1: Self-confidence (~$0.001/test)
The analysis model reports its own confidence score. If ≥95%, the result passes without further verification. ~80% of results resolve here.
Stage 2: Self-consistency (~$0.003/test)
The same model re-evaluates the result 3 times with different prompts. If all 3 agree, the result is confirmed. Catches inconsistent reasoning.
Stage 3: Cross-model vote (~$0.01/test)
Multiple model tiers evaluate the result independently. A majority vote decides. Catches model-specific biases.
Stage 4: Structured debate (~$0.05/test)
Two model instances argue opposing conclusions with structured arguments. A judge model evaluates the debate and renders a final decision. Full audit trail preserved.
MVP status
Phase 1 uses Stage 1 only (confidence threshold). The full cascade is planned for Phase 2.
Target metrics
| Metric | Target |
|---|---|
| False Positive Rate (FPR) | < 5% |
| Stage 1 resolution rate | ~80% of results |
| Average cost per verification | < $0.005 |
What's next
- 5-layer deep dive — Full pipeline architecture
- Model orchestration — Model tier routing