GLINER detector
nuner-zero PII detection benchmark
Measured on 41 of 41 retained datasets under the same frozen scoring and masking protocol.
71.63%fully hidden annotations
78.78%detected / overlapped
3.94%extra masking
497CPU characters / second
Strengths
Highest complete-masking rates by normalized data type.
Network identifiers96.24%24,667 / 25,631
Logins & usernames83.19%9,469 / 11,382
Phone numbers & email79.94%17,572 / 21,981
Watch list
Lowest observed complete-masking rates; inspect before deployment.
Other sensitive attributes43.31%1,389 / 3,207
Documents & identifiers60.59%9,109 / 15,034
Bank accounts & cards62.73%3,295 / 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 | 75.02% | 79.19% | 4.51% |
| Russian | 19 | 71.76% | 80.80% | 2.70% |
| Multilingual | 5 | 68.91% | 73.35% | 2.35% |
Configuration, source and version
- Family
- GLINER
- Revision
- c90187673f464518dca09f41689184ed6976242c
- Eligible datasets
- 41 / 41
- Character F1
- 0.760
- Untouched diagnostic
- 21.22% · 48,263 / 227,466
- Benchmark version
- 1.0.3 · 2026-09-09
- CPU seconds / 10k chars
- 20.121
- GPU seconds / 10k chars
- 0.382
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 | 7.55% | 95.37% | — |
| alexen2 | 1,261 | 27.91% | 99.21% | 11.27% |
| synth-env-configs | 392 | 37.50% | 41.07% | 9.37% |
| corrupt-secrets-issues | 288 | 40.97% | 70.49% | 2.23% |
| multiconer-ru | 1,208 | 43.63% | 61.34% | 9.66% |
| corrupt-redmadrobot | 5,531 | 45.98% | 55.45% | 0.73% |
| secrets-issues | 288 | 49.31% | 70.49% | 2.53% |
| leaky-repo | 95 | 50.53% | 61.05% | 3.08% |
| synth-secrets-ru | 581 | 52.15% | 65.92% | 5.14% |
| synth-secrets-en | 581 | 53.18% | 64.37% | 4.00% |
| factrueval | 7,966 | 53.57% | 54.81% | — |
| nym-ru | 9,497 | 58.63% | 65.92% | — |
| nym-multi | 8,169 | 59.29% | 63.52% | 4.31% |
| redmadrobot | 5,516 | 62.20% | 65.05% | 0.54% |
| dialogpii-multi | 10,345 | 62.71% | 68.20% | — |
| synth-wiki-tables | 13,094 | 64.04% | 78.87% | — |
| leak-museum | 101 | 64.36% | 70.30% | 5.14% |
| creddata | 776 | 64.82% | 74.10% | 6.43% |
| secrets-rules | 746 | 65.68% | 72.52% | 37.52% |
| corrupt-hivetrace | 1,667 | 67.25% | 84.28% | 2.07% |
| tonicai | 2,417 | 67.48% | 71.49% | 5.26% |
| nym-en | 3,584 | 68.61% | 70.90% | 3.46% |
| gretel-multi | 4,484 | 69.11% | 73.57% | 2.32% |
| jayguard | 1,195 | 71.05% | 84.69% | 1.71% |
| redact-multi | 22,864 | 73.43% | 78.46% | — |
| nerel | 24,369 | 73.61% | 78.89% | — |
| kiji-en | 7,623 | 73.68% | 76.16% | — |
| redact-ru | 9,230 | 73.86% | 82.17% | 12.84% |
| kiji-multi | 7,565 | 74.04% | 75.45% | — |
| scanpatch | 8,708 | 76.22% | 82.44% | 7.76% |
| privy | 1,899 | 76.67% | 79.99% | 2.17% |
| alrosait | 1,862 | 77.93% | 99.36% | 10.18% |
| nemotron-pii | 9,391 | 78.33% | 80.13% | 2.13% |
| synth-jira-comments | 2,705 | 79.00% | 82.07% | — |
| dialogpii-en | 3,087 | 80.24% | 85.36% | 12.29% |
| russian-pii-66k | 4,805 | 82.14% | 86.58% | — |
| hivetrace | 1,667 | 82.72% | 88.12% | 2.52% |
| ameau01 | 2,242 | 84.34% | 84.52% | 6.68% |
| tab-echr | 3,830 | 84.60% | 96.14% | — |
| synth-ru-tickets | 32,099 | 93.41% | 96.07% | — |
| arthur-passwords | 280 | 97.50% | 97.50% | 4.53% |