GLINER detector
gliner-pii-base PII detection benchmark
Measured on 41 of 41 retained datasets under the same frozen scoring and masking protocol.
44.64%fully hidden annotations
50.53%detected / overlapped
1.86%extra masking
1,907CPU characters / second
Strengths
Highest complete-masking rates by normalized data type.
Phone numbers & email66.87%14,698 / 21,981
Logins & usernames60.78%6,918 / 11,382
Organizations54.23%10,453 / 19,274
Watch list
Lowest observed complete-masking rates; inspect before deployment.
Other sensitive attributes20.42%655 / 3,207
Passwords, keys & tokens33.48%2,300 / 6,870
People's names34.00%22,352 / 65,736
Language coverage
The overall leader can differ from the best choice for one language slice.
| Language | Datasets | Fully hidden ↑ | Detected ↑ | Extra masking ↓ |
|---|---|---|---|---|
| English | 17 | 63.07% | 65.86% | 2.11% |
| Russian | 19 | 41.97% | 49.33% | 1.35% |
| Multilingual | 5 | 38.50% | 42.79% | 0.69% |
Configuration, source and version
- Family
- GLINER
- Revision
- 61726e0ad791dcab3e29339bbec3ad42ded65641
- Eligible datasets
- 41 / 41
- Character F1
- 0.550
- Untouched diagnostic
- 49.47% · 112,527 / 227,466
- Benchmark version
- 1.0.3 · 2026-09-09
- CPU seconds / 10k chars
- 5.244
- GPU seconds / 10k chars
- 0.157
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 | 6.13% | 63.79% | — |
| corrupt-secrets-issues | 288 | 7.64% | 13.54% | 0.56% |
| secrets-issues | 288 | 9.38% | 14.58% | 0.72% |
| leaky-repo | 95 | 11.58% | 14.74% | 0.62% |
| leak-museum | 101 | 16.83% | 16.83% | 0.96% |
| creddata | 776 | 17.40% | 18.94% | 2.82% |
| synth-env-configs | 392 | 18.11% | 20.66% | 3.26% |
| secrets-rules | 746 | 20.78% | 22.12% | 19.04% |
| synth-secrets-ru | 581 | 21.51% | 23.41% | 1.91% |
| alexen2 | 1,261 | 24.43% | 94.53% | 10.62% |
| scanpatch | 8,708 | 26.19% | 32.46% | 3.29% |
| multiconer-ru | 1,208 | 26.24% | 40.23% | 5.47% |
| redact-ru | 9,230 | 27.41% | 35.73% | 3.41% |
| redact-multi | 22,864 | 27.93% | 33.63% | — |
| corrupt-hivetrace | 1,667 | 29.57% | 31.85% | 0.78% |
| synth-secrets-en | 581 | 31.15% | 34.60% | 2.63% |
| synth-wiki-tables | 13,094 | 31.86% | 32.97% | — |
| corrupt-redmadrobot | 5,531 | 32.67% | 36.16% | 0.60% |
| nym-ru | 9,497 | 35.11% | 39.72% | — |
| nym-multi | 8,169 | 39.11% | 40.34% | 0.75% |
| gretel-multi | 4,484 | 39.12% | 45.81% | 0.69% |
| russian-pii-66k | 4,805 | 39.19% | 42.58% | — |
| tonicai | 2,417 | 39.35% | 42.28% | 2.29% |
| factrueval | 7,966 | 40.70% | 41.99% | — |
| hivetrace | 1,667 | 41.03% | 42.77% | 0.84% |
| jayguard | 1,195 | 43.68% | 63.18% | 1.60% |
| redmadrobot | 5,516 | 44.14% | 47.75% | 0.53% |
| synth-jira-comments | 2,705 | 45.80% | 47.28% | — |
| kiji-multi | 7,565 | 49.19% | 50.01% | — |
| synth-ru-tickets | 32,099 | 49.48% | 57.47% | — |
| nym-en | 3,584 | 50.36% | 50.61% | 2.31% |
| alrosait | 1,862 | 51.07% | 65.41% | 3.37% |
| dialogpii-multi | 10,345 | 53.31% | 58.38% | — |
| privy | 1,899 | 57.56% | 58.29% | 3.08% |
| nerel | 24,369 | 60.88% | 66.06% | — |
| kiji-en | 7,623 | 62.06% | 63.64% | — |
| dialogpii-en | 3,087 | 71.59% | 80.01% | 5.15% |
| ameau01 | 2,242 | 75.69% | 75.74% | 4.74% |
| tab-echr | 3,830 | 77.65% | 83.29% | — |
| nemotron-pii | 9,391 | 78.46% | 81.34% | 1.09% |
| arthur-passwords | 280 | 100.00% | 100.00% | 4.63% |