Development mock and self-reported results are not official. Official results require an organizer-run private-holdout evaluation.
SanctionScore is offending-authority recall on a 0 to 100 scale. Clean audits are documents with no missed offending authorities. FPR and control alarms measure incorrect flags against real authorities or clean control documents. Extra reviews per document counts flags that require additional verification.
Scroll horizontally to view all metrics.
| Tier | Model | Revision | Organic closed | Organic clean audits | Control alarms | Organic extra reviews/doc | Citation closed | Closed FPR | Citation tool | Tool FPR | Document closed | Document tool | Tool clean audits | Tool extra reviews/doc | Queries |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| development_mock | mock/deterministic-rule-based-v1 | deterministic-rule-based-v1 | N/A | N/A | N/A | N/A | 41.67 | 33.33% | 100.00 | 8.33% | 75.00 | 100.00 | 100.00% | 1.00 | 50 |