Permission fidelity
Compare disclosure decisions against source-native access and purpose rules.
Benchmarks
Measure permission fidelity, temporal correctness, sufficiency, and replay.
Decision-evidence systems need benchmarks that test whether the right evidence is admitted—and whether forbidden, stale, contradictory, or missing evidence is handled correctly.
Benchmark design · planned
Temporal accuracy
Test valid-time and known-at reconstruction across change sequences.
Replay fidelity
Measure exact, proven, equivalent, partial, and failed reconstruction.
Current limitationNo performance claims are published until repeatable methodology and vectors are available.