GLINER2 detector
gliner2-large PII detection benchmark
Measured on 41 of 41 retained datasets under the same frozen scoring and masking protocol.
65.65%fully hidden annotations
71.98%detected / overlapped
3.51%extra masking
405CPU characters / second
Strengths
Highest complete-masking rates by normalized data type.
Phone numbers & email87.70%19,277 / 21,981
Logins & usernames82.60%9,402 / 11,382
People's names67.46%44,344 / 65,736
Watch list
Lowest observed complete-masking rates; inspect before deployment.
Passwords, keys & tokens40.09%2,754 / 6,870
Other sensitive attributes41.47%1,330 / 3,207
Dates & times48.56%2,672 / 5,502
Language coverage
The overall leader can differ from the best choice for one language slice.
| Language | Datasets | Fully hidden ↑ | Detected ↑ | Extra masking ↓ |
|---|---|---|---|---|
| English | 17 | 77.50% | 80.94% | 3.72% |
| Russian | 19 | 63.79% | 71.50% | 3.05% |
| Multilingual | 5 | 62.06% | 66.90% | 3.02% |
Configuration, source and version
- Family
- GLINER2
- Revision
- 6a498b5a28ec3908bbc5277aeb47d22bcfc02f33
- Eligible datasets
- 41 / 41
- Character F1
- 0.685
- Untouched diagnostic
- 28.02% · 63,732 / 227,466
- Benchmark version
- 1.0.3 · 2026-09-09
- CPU seconds / 10k chars
- 24.691
- GPU seconds / 10k chars
- 0.607
Results by dataset
Complete masking is primary; detection shows whether any character overlapped.
| Dataset | Gold annotations | Fully hidden ↑ | Detected ↑ | Extra masking ↓ |
|---|---|---|---|---|
| secrets-rules | 746 | 4.29% | 5.36% | 20.69% |
| rubai-ru | 3,458 | 6.45% | 89.59% | — |
| synth-env-configs | 392 | 7.40% | 7.40% | 5.54% |
| corrupt-secrets-issues | 288 | 11.46% | 29.86% | 2.04% |
| secrets-issues | 288 | 13.54% | 32.29% | 2.31% |
| synth-secrets-en | 581 | 17.73% | 19.45% | 2.10% |
| alexen2 | 1,261 | 21.73% | 93.50% | 6.57% |
| synth-secrets-ru | 581 | 23.75% | 26.68% | 3.45% |
| leaky-repo | 95 | 32.63% | 41.05% | 3.45% |
| leak-museum | 101 | 40.59% | 44.55% | 6.04% |
| creddata | 776 | 44.07% | 47.81% | 6.76% |
| scanpatch | 8,708 | 48.97% | 54.96% | 7.18% |
| corrupt-hivetrace | 1,667 | 49.85% | 59.93% | 2.50% |
| corrupt-redmadrobot | 5,531 | 50.23% | 55.58% | 1.05% |
| alrosait | 1,862 | 51.18% | 87.92% | 10.94% |
| nym-multi | 8,169 | 52.85% | 55.34% | 0.00% |
| dialogpii-multi | 10,345 | 54.87% | 60.68% | — |
| multiconer-ru | 1,208 | 54.97% | 74.25% | 17.37% |
| redact-ru | 9,230 | 58.33% | 66.84% | 6.96% |
| jayguard | 1,195 | 58.58% | 73.39% | 2.44% |
| nym-ru | 9,497 | 59.01% | 63.30% | — |
| hivetrace | 1,667 | 59.21% | 66.95% | 3.33% |
| redmadrobot | 5,516 | 61.80% | 65.36% | 1.07% |
| redact-multi | 22,864 | 62.34% | 67.31% | — |
| privy | 1,899 | 64.56% | 68.62% | 4.31% |
| gretel-multi | 4,484 | 65.77% | 77.68% | 3.06% |
| nerel | 24,369 | 66.55% | 70.47% | — |
| nym-en | 3,584 | 66.96% | 70.45% | 0.54% |
| factrueval | 7,966 | 70.60% | 71.55% | — |
| synth-jira-comments | 2,705 | 70.72% | 73.64% | — |
| synth-ru-tickets | 32,099 | 73.75% | 79.31% | — |
| synth-wiki-tables | 13,094 | 74.23% | 75.55% | — |
| ameau01 | 2,242 | 76.14% | 76.23% | 6.03% |
| russian-pii-66k | 4,805 | 76.15% | 77.50% | — |
| dialogpii-en | 3,087 | 78.00% | 85.84% | 9.92% |
| kiji-multi | 7,565 | 78.84% | 80.24% | — |
| tab-echr | 3,830 | 83.42% | 93.71% | — |
| kiji-en | 7,623 | 85.92% | 86.72% | — |
| arthur-passwords | 280 | 86.07% | 86.07% | 2.79% |
| nemotron-pii | 9,391 | 91.18% | 92.92% | 0.36% |
| tonicai | 2,417 | 91.64% | 94.37% | 5.65% |