Policy Performance
Stress-tested correctness profile for domain policy decisions.
Stress-tested correctness profile for domain policy decisions.
50-100 edge cases per domain with block/allow/override expectation matching.
| Domain | Cases | Allow | Warn | Challenge | Block | Override | Override rate | False + | False - | Accuracy |
|---|---|---|---|---|---|---|---|---|---|---|
| labour | 75 | 61 | 0 | 0 | 14 | 0 | 0.00% | 6 | 42 | 36.00% |
| legal | 75 | 50 | 0 | 0 | 25 | 0 | 0.00% | 0 | 25 | 66.67% |
| commerce | 75 | 50 | 0 | 0 | 25 | 0 | 0.00% | 0 | 25 | 66.67% |
| media-rights | 75 | 50 | 0 | 25 | 0 | 25 | 33.33% | 0 | 25 | 66.67% |
Overall accuracy: 59.00% · false positives: 6 · false negatives: 117