OPF detector
opf-kz-ru PII detection benchmark
Measured on 41 of 41 retained datasets under the same frozen scoring and masking protocol.
56.51%fully hidden annotations
65.16%detected / overlapped
10.90%extra masking
2,950CPU characters / second
Strengths
Highest complete-masking rates by normalized data type.
Phone numbers & email89.87%19,755 / 21,981
Dates & times89.71%4,936 / 5,502
Passwords, keys & tokens89.45%6,145 / 6,870
Watch list
Lowest observed complete-masking rates; inspect before deployment.
Organizations4.05%780 / 19,274
Other sensitive attributes12.60%404 / 3,207
Addresses & locations32.16%12,735 / 39,602
Language coverage
The overall leader can differ from the best choice for one language slice.
| Language | Datasets | Fully hidden ↑ | Detected ↑ | Extra masking ↓ |
|---|---|---|---|---|
| English | 17 | 66.67% | 70.15% | 11.79% |
| Russian | 19 | 52.65% | 64.12% | 9.22% |
| Multilingual | 5 | 59.18% | 64.29% | 1.56% |
Configuration, source and version
- Family
- OPF
- Revision
- b804356ff8ac827fe2b046e8d3d0be2720c6b148
- Eligible datasets
- 41 / 41
- Character F1
- 0.634
- Untouched diagnostic
- 34.84% · 79,257 / 227,466
- Benchmark version
- 1.0.3 · 2026-09-09
- CPU seconds / 10k chars
- 3.390
- GPU seconds / 10k chars
- 1.594
Results by dataset
Complete masking is primary; detection shows whether any character overlapped.
| Dataset | Gold annotations | Fully hidden ↑ | Detected ↑ | Extra masking ↓ |
|---|---|---|---|---|
| multiconer-ru | 1,208 | 2.15% | 4.06% | 1.62% |
| rubai-ru | 3,458 | 4.22% | 93.00% | — |
| nerel | 24,369 | 9.40% | 11.80% | — |
| factrueval | 7,966 | 15.03% | 15.35% | — |
| alexen2 | 1,261 | 21.89% | 92.39% | 13.46% |
| tab-echr | 3,830 | 23.32% | 26.79% | — |
| dialogpii-multi | 10,345 | 24.65% | 29.55% | — |
| jayguard | 1,195 | 26.69% | 44.35% | 2.35% |
| dialogpii-en | 3,087 | 43.34% | 48.36% | 6.39% |
| alrosait | 1,862 | 44.04% | 84.96% | 4.07% |
| ameau01 | 2,242 | 50.80% | 53.79% | 4.04% |
| tonicai | 2,417 | 53.37% | 56.60% | 0.21% |
| corrupt-redmadrobot | 5,531 | 54.28% | 67.44% | 3.25% |
| scanpatch | 8,708 | 59.30% | 68.73% | 1.90% |
| gretel-multi | 4,484 | 61.71% | 70.25% | 1.54% |
| redact-multi | 22,864 | 63.37% | 69.30% | — |
| redmadrobot | 5,516 | 64.54% | 72.77% | 2.26% |
| synth-jira-comments | 2,705 | 67.65% | 76.38% | — |
| privy | 1,899 | 68.51% | 75.09% | 10.21% |
| synth-ru-tickets | 32,099 | 68.95% | 82.99% | — |
| leak-museum | 101 | 70.30% | 74.26% | 3.74% |
| nemotron-pii | 9,391 | 71.39% | 74.44% | 2.55% |
| redact-ru | 9,230 | 72.24% | 79.26% | 8.42% |
| corrupt-hivetrace | 1,667 | 73.25% | 87.04% | 2.52% |
| corrupt-secrets-issues | 288 | 73.26% | 95.49% | 8.23% |
| nym-ru | 9,497 | 73.87% | 81.59% | — |
| nym-multi | 8,169 | 74.17% | 78.74% | 3.18% |
| leaky-repo | 95 | 75.79% | 83.16% | 3.83% |
| kiji-multi | 7,565 | 76.07% | 77.55% | — |
| secrets-issues | 288 | 76.39% | 96.88% | 7.23% |
| synth-wiki-tables | 13,094 | 79.37% | 88.59% | — |
| russian-pii-66k | 4,805 | 79.65% | 86.56% | — |
| nym-en | 3,584 | 81.64% | 85.13% | 2.81% |
| kiji-en | 7,623 | 82.82% | 83.75% | — |
| synth-env-configs | 392 | 83.16% | 99.74% | 22.84% |
| hivetrace | 1,667 | 83.80% | 92.50% | 0.90% |
| synth-secrets-en | 581 | 91.39% | 98.28% | 47.61% |
| synth-secrets-ru | 581 | 94.15% | 99.31% | 49.04% |
| creddata | 776 | 95.36% | 97.94% | 6.61% |
| arthur-passwords | 280 | 97.50% | 97.50% | 15.29% |
| secrets-rules | 746 | 98.39% | 99.33% | 51.19% |