GLINER2 detector
gliner25-fastino PII detection benchmark
Measured on 41 of 41 retained datasets under the same frozen scoring and masking protocol.
62.77%fully hidden annotations
69.13%detected / overlapped
7.33%extra masking
802CPU characters / second
Strengths
Highest complete-masking rates by normalized data type.
Phone numbers & email91.01%20,005 / 21,981
People's names75.80%49,828 / 65,736
Addresses & locations64.60%25,584 / 39,602
Watch list
Lowest observed complete-masking rates; inspect before deployment.
Customer & employee IDs16.42%1,313 / 7,994
Dates & times22.81%1,255 / 5,502
Other sensitive attributes25.26%810 / 3,207
Language coverage
The overall leader can differ from the best choice for one language slice.
| Language | Datasets | Fully hidden ↑ | Detected ↑ | Extra masking ↓ |
|---|---|---|---|---|
| English | 17 | 70.05% | 73.15% | 8.08% |
| Russian | 19 | 60.99% | 69.37% | 5.87% |
| Multilingual | 5 | 62.17% | 65.69% | 0.75% |
Configuration, source and version
- Family
- GLINER2
- Revision
- aaecfe45db1d828c963717054ccb868e8ad1f1d5
- Eligible datasets
- 41 / 41
- Character F1
- 0.696
- Untouched diagnostic
- 30.87% · 70,219 / 227,466
- Benchmark version
- 1.0.3 · 2026-09-09
- CPU seconds / 10k chars
- 12.469
- GPU seconds / 10k chars
- 0.153
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 | 7.92% | 97.95% | — |
| corrupt-secrets-issues | 288 | 23.26% | 42.36% | 5.64% |
| secrets-issues | 288 | 24.31% | 40.97% | 5.40% |
| alexen2 | 1,261 | 26.96% | 99.60% | 15.90% |
| corrupt-redmadrobot | 5,531 | 32.92% | 36.61% | 1.50% |
| nym-multi | 8,169 | 39.58% | 41.45% | 0.00% |
| nym-ru | 9,497 | 41.86% | 44.66% | — |
| redmadrobot | 5,516 | 42.22% | 44.02% | 1.59% |
| corrupt-hivetrace | 1,667 | 47.21% | 53.51% | 0.98% |
| nym-en | 3,584 | 49.30% | 51.48% | 0.63% |
| synth-env-configs | 392 | 49.49% | 50.26% | 12.58% |
| leak-museum | 101 | 52.48% | 54.46% | 8.11% |
| leaky-repo | 95 | 53.68% | 64.21% | 16.01% |
| synth-secrets-en | 581 | 54.04% | 56.80% | 22.91% |
| hivetrace | 1,667 | 55.79% | 58.19% | 1.14% |
| factrueval | 7,966 | 55.96% | 56.69% | — |
| synth-secrets-ru | 581 | 56.80% | 60.07% | 24.40% |
| dialogpii-multi | 10,345 | 59.15% | 64.02% | — |
| multiconer-ru | 1,208 | 60.93% | 80.30% | 16.03% |
| scanpatch | 8,708 | 61.04% | 65.14% | 3.00% |
| privy | 1,899 | 61.19% | 63.03% | 2.77% |
| synth-wiki-tables | 13,094 | 63.78% | 65.27% | — |
| russian-pii-66k | 4,805 | 63.89% | 66.24% | — |
| synth-ru-tickets | 32,099 | 66.65% | 79.04% | — |
| redact-multi | 22,864 | 67.60% | 71.45% | — |
| redact-ru | 9,230 | 68.68% | 73.58% | 5.53% |
| dialogpii-en | 3,087 | 69.00% | 75.96% | 8.12% |
| kiji-multi | 7,565 | 69.31% | 70.28% | — |
| nemotron-pii | 9,391 | 70.39% | 72.30% | 0.36% |
| gretel-multi | 4,484 | 70.56% | 76.61% | 0.76% |
| tonicai | 2,417 | 72.45% | 73.98% | 2.65% |
| ameau01 | 2,242 | 72.48% | 72.52% | 2.75% |
| nerel | 24,369 | 74.09% | 77.69% | — |
| kiji-en | 7,623 | 74.09% | 75.00% | — |
| jayguard | 1,195 | 74.48% | 86.44% | 1.85% |
| synth-jira-comments | 2,705 | 81.11% | 81.89% | — |
| creddata | 776 | 82.09% | 86.34% | 10.69% |
| tab-echr | 3,830 | 86.84% | 96.01% | — |
| alrosait | 1,862 | 87.06% | 99.57% | 10.77% |
| secrets-rules | 746 | 89.95% | 94.24% | 40.18% |
| arthur-passwords | 280 | 98.93% | 98.93% | 7.25% |