GLINER detector
gliner-urchade PII detection benchmark
Measured on 41 of 41 retained datasets under the same frozen scoring and masking protocol.
65.51%fully hidden annotations
72.52%detected / overlapped
3.79%extra masking
1,033CPU characters / second
Strengths
Highest complete-masking rates by normalized data type.
Logins & usernames89.11%10,142 / 11,382
Phone numbers & email80.40%17,672 / 21,981
Organizations71.82%13,842 / 19,274
Watch list
Lowest observed complete-masking rates; inspect before deployment.
Network identifiers46.46%11,908 / 25,631
Other sensitive attributes48.02%1,540 / 3,207
Customer & employee IDs53.77%4,298 / 7,994
Language coverage
The overall leader can differ from the best choice for one language slice.
| Language | Datasets | Fully hidden ↑ | Detected ↑ | Extra masking ↓ |
|---|---|---|---|---|
| English | 17 | 62.54% | 65.18% | 4.42% |
| Russian | 19 | 63.55% | 73.53% | 2.38% |
| Multilingual | 5 | 72.62% | 75.09% | 2.75% |
Configuration, source and version
- Family
- GLINER
- Revision
- 1fcf13e85f4eef5394e1fcd406cf2ca9ea82351d
- Eligible datasets
- 41 / 41
- Character F1
- 0.687
- Untouched diagnostic
- 27.48% · 62,517 / 227,466
- Benchmark version
- 1.0.3 · 2026-09-09
- CPU seconds / 10k chars
- 9.681
- GPU seconds / 10k chars
- 0.203
Results by dataset
Complete masking is primary; detection shows whether any character overlapped.
| Dataset | Gold annotations | Fully hidden ↑ | Detected ↑ | Extra masking ↓ |
|---|---|---|---|---|
| rubai-ru | 3,458 | 7.17% | 78.48% | — |
| synth-env-configs | 392 | 27.81% | 27.81% | 5.36% |
| synth-secrets-en | 581 | 27.88% | 30.64% | 1.31% |
| alexen2 | 1,261 | 27.99% | 99.44% | 7.08% |
| synth-secrets-ru | 581 | 32.01% | 34.94% | 1.63% |
| corrupt-secrets-issues | 288 | 34.38% | 55.21% | 2.86% |
| tonicai | 2,417 | 37.73% | 42.20% | 2.93% |
| factrueval | 7,966 | 40.07% | 40.58% | — |
| leaky-repo | 95 | 42.11% | 51.58% | 5.70% |
| secrets-issues | 288 | 43.75% | 61.81% | 2.72% |
| leak-museum | 101 | 45.54% | 49.50% | 9.10% |
| secrets-rules | 746 | 46.65% | 50.54% | 27.53% |
| nym-en | 3,584 | 47.63% | 48.05% | 2.68% |
| nym-multi | 8,169 | 53.32% | 54.40% | 0.00% |
| corrupt-redmadrobot | 5,531 | 53.35% | 55.96% | 1.06% |
| nym-ru | 9,497 | 56.28% | 58.86% | — |
| creddata | 776 | 57.99% | 62.37% | 8.74% |
| privy | 1,899 | 59.29% | 60.45% | 5.23% |
| redmadrobot | 5,516 | 60.10% | 61.53% | 0.96% |
| kiji-en | 7,623 | 60.13% | 61.59% | — |
| synth-wiki-tables | 13,094 | 62.22% | 64.43% | — |
| corrupt-hivetrace | 1,667 | 62.57% | 66.71% | 2.01% |
| synth-ru-tickets | 32,099 | 64.00% | 85.94% | — |
| multiconer-ru | 1,208 | 66.31% | 73.34% | 11.50% |
| scanpatch | 8,708 | 66.69% | 72.57% | 7.05% |
| synth-jira-comments | 2,705 | 67.13% | 68.72% | — |
| kiji-multi | 7,565 | 67.36% | 68.27% | — |
| nemotron-pii | 9,391 | 69.91% | 72.19% | 0.00% |
| hivetrace | 1,667 | 70.31% | 71.21% | 2.21% |
| russian-pii-66k | 4,805 | 70.36% | 71.49% | — |
| dialogpii-multi | 10,345 | 70.47% | 75.07% | — |
| gretel-multi | 4,484 | 73.75% | 76.09% | 2.79% |
| nerel | 24,369 | 75.37% | 78.86% | — |
| dialogpii-en | 3,087 | 75.38% | 81.54% | 13.79% |
| jayguard | 1,195 | 76.40% | 86.86% | 2.37% |
| tab-echr | 3,830 | 76.87% | 79.69% | — |
| ameau01 | 2,242 | 77.16% | 78.06% | 5.14% |
| redact-ru | 9,230 | 80.55% | 87.02% | 16.19% |
| redact-multi | 22,864 | 82.02% | 84.54% | — |
| alrosait | 1,862 | 90.28% | 92.43% | 8.26% |
| arthur-passwords | 280 | 90.36% | 90.36% | 2.29% |