CustodianAI · PHI Benchmark Dashboard

Overall results · 6 systems × all benchmarks

F1 per benchmark plus mean F1. Rows sorted by overall mean. Greener = better.

Metric:

Ranking proof · one document, six systems side by side

Same input text, predictions overlaid. Sorted by this document's F1 — top panel is the best system on this doc, bottom panel is the worst. green = correctly caught PHI, yellow = false positive, red = missed PHI (leakage).

Benchmark:

Per-benchmark detail view (click to expand)

Benchmark

Matching mode

Highlight metric

Per-system comparison

Click a column header to sort. Rows highlighted by the selected metric.

System	P	R	F1	Leak	Char leak	n

Precision vs Recall

Leakage rate (lower = better)

Per-document drill-down

Sampled docs where systems disagree most. Hover a span to see its label.