Product scores
Pending real-machine measurement
When a real run exists, its scorer --out-json report lands in runs[]
and this page will show, per product and per run:
- Detection rate — signal-class primary column + any-hit fallback, by class and tactic.
- False-positive rate — verified / attempt tiers (calibre v2.2).
- Per-class hit heatmap and median latency (attribution-window calibre, not true MTTD).
- Multi-product comparison on the same environment and case set.
Every number carries its run metadata (corpus sha, calibre, target, timestamp). Any synthetic data used for UI debugging is labelled SYNTHETIC DEMO.