GLINER detector
gliner-multi-v21 PII detection benchmark
Measured on 41 of 41 retained datasets under the same frozen scoring and masking protocol.
51.28%fully hidden annotations
60.10%detected / overlapped
2.71%extra masking
1,011CPU characters / second
Strengths
Highest complete-masking rates by normalized data type.
People's names82.42%54,179 / 65,736
Logins & usernames67.89%7,727 / 11,382
Organizations61.99%11,948 / 19,274
Watch list
Lowest observed complete-masking rates; inspect before deployment.
Customer & employee IDs10.20%815 / 7,994
Network identifiers16.30%4,177 / 25,631
Bank accounts & cards18.24%958 / 5,253
Language coverage
The overall leader can differ from the best choice for one language slice.
| Language | Datasets | Fully hidden ↑ | Detected ↑ | Extra masking ↓ |
|---|---|---|---|---|
| English | 17 | 58.52% | 64.17% | 2.90% |
| Russian | 19 | 46.43% | 57.34% | 2.31% |
| Multilingual | 5 | 58.57% | 64.28% | 1.90% |
Configuration, source and version
- Family
- GLINER
- Revision
- 443d26d654e0324125a96bebd8e796c14ff2efe6
- Eligible datasets
- 41 / 41
- Character F1
- 0.537
- Untouched diagnostic
- 39.90% · 90,767 / 227,466
- Benchmark version
- 1.0.3 · 2026-09-09
- CPU seconds / 10k chars
- 9.891
- GPU seconds / 10k chars
- 0.186
Results by dataset
Complete masking is primary; detection shows whether any character overlapped.
| Dataset | Gold annotations | Fully hidden ↑ | Detected ↑ | Extra masking ↓ |
|---|---|---|---|---|
| secrets-rules | 746 | 1.47% | 3.22% | 10.36% |
| rubai-ru | 3,458 | 6.04% | 86.41% | — |
| corrupt-secrets-issues | 288 | 8.33% | 26.39% | 1.88% |
| secrets-issues | 288 | 11.81% | 29.51% | 2.07% |
| synth-secrets-ru | 581 | 14.46% | 19.97% | 1.57% |
| synth-wiki-tables | 13,094 | 18.14% | 22.34% | — |
| synth-secrets-en | 581 | 18.24% | 25.13% | 1.95% |
| creddata | 776 | 20.36% | 24.48% | 4.52% |
| synth-env-configs | 392 | 20.66% | 23.21% | 4.45% |
| leaky-repo | 95 | 22.11% | 30.53% | 1.58% |
| leak-museum | 101 | 22.77% | 28.71% | 3.67% |
| corrupt-redmadrobot | 5,531 | 25.04% | 29.04% | 0.75% |
| alexen2 | 1,261 | 26.96% | 87.55% | 6.12% |
| corrupt-hivetrace | 1,667 | 28.01% | 44.15% | 2.74% |
| redmadrobot | 5,516 | 35.57% | 39.38% | 0.95% |
| nym-multi | 8,169 | 37.42% | 39.99% | 0.00% |
| nym-ru | 9,497 | 37.50% | 42.96% | — |
| hivetrace | 1,667 | 37.85% | 52.07% | 3.04% |
| synth-jira-comments | 2,705 | 38.08% | 44.88% | — |
| nym-en | 3,584 | 38.11% | 40.29% | 0.26% |
| synth-ru-tickets | 32,099 | 38.43% | 54.35% | — |
| privy | 1,899 | 39.23% | 43.55% | 5.57% |
| multiconer-ru | 1,208 | 48.84% | 66.06% | 8.16% |
| scanpatch | 8,708 | 53.55% | 59.62% | 3.41% |
| gretel-multi | 4,484 | 53.86% | 63.31% | 1.93% |
| russian-pii-66k | 4,805 | 54.80% | 65.20% | — |
| redact-ru | 9,230 | 56.88% | 67.63% | 11.35% |
| nemotron-pii | 9,391 | 59.01% | 65.91% | 0.61% |
| redact-multi | 22,864 | 60.56% | 67.84% | — |
| ameau01 | 2,242 | 62.00% | 64.63% | 2.51% |
| dialogpii-multi | 10,345 | 64.06% | 70.09% | — |
| tonicai | 2,417 | 66.16% | 82.13% | 2.43% |
| kiji-en | 7,623 | 67.32% | 69.70% | — |
| kiji-multi | 7,565 | 70.69% | 72.37% | — |
| alrosait | 1,862 | 71.64% | 99.52% | 10.82% |
| nerel | 24,369 | 72.27% | 76.46% | — |
| jayguard | 1,195 | 73.22% | 86.78% | 2.48% |
| dialogpii-en | 3,087 | 75.06% | 82.38% | 12.48% |
| factrueval | 7,966 | 75.40% | 76.26% | — |
| tab-echr | 3,830 | 83.63% | 90.21% | — |
| arthur-passwords | 280 | 94.64% | 94.64% | 5.49% |