GLINER detector
gliner-nvidia PII detection benchmark
Measured on 40 of 41 retained datasets under the same frozen scoring and masking protocol.
63.97%fully hidden annotations
74.53%detected / overlapped
4.92%extra masking
442CPU characters / second
Strengths
Highest complete-masking rates by normalized data type.
Logins & usernames74.99%8,364 / 11,153
Phone numbers & email74.83%15,552 / 20,782
Dates & times73.26%3,828 / 5,225
Watch list
Lowest observed complete-masking rates; inspect before deployment.
Other sensitive attributes36.83%1,181 / 3,207
Network identifiers42.31%10,521 / 24,869
Organizations55.57%10,242 / 18,431
Language coverage
The overall leader can differ from the best choice for one language slice.
| Language | Datasets | Fully hidden ↑ | Detected ↑ | Extra masking ↓ |
|---|---|---|---|---|
| English | 16 | 79.30% | 85.22% | 5.35% |
| Russian | 19 | 59.74% | 69.87% | 4.02% |
| Multilingual | 5 | 66.65% | 80.80% | 2.60% |
Configuration, source and version
- Family
- GLINER
- Revision
- bd23e8ef4425fd04e34c5204ab49ffaa706eae79
- Eligible datasets
- 40 / 41
- Character F1
- 0.677
- Untouched diagnostic
- 25.47% · 55,538 / 218,075
- Benchmark version
- 1.0.3 · 2026-09-09
- CPU seconds / 10k chars
- 22.624
- GPU seconds / 10k chars
- 0.372
source overlap: nemotron-pii
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 | 7.17% | 80.36% | — |
| alexen2 | 1,261 | 26.49% | 98.57% | 6.31% |
| corrupt-secrets-issues | 288 | 29.86% | 47.57% | 3.28% |
| secrets-issues | 288 | 40.63% | 57.99% | 3.19% |
| multiconer-ru | 1,208 | 44.04% | 58.94% | 10.13% |
| synth-env-configs | 392 | 44.13% | 45.15% | 6.60% |
| leaky-repo | 95 | 49.47% | 56.84% | 0.59% |
| synth-ru-tickets | 32,099 | 50.44% | 53.72% | — |
| gretel-multi | 4,484 | 50.87% | 83.07% | 2.52% |
| synth-wiki-tables | 13,094 | 53.50% | 61.65% | — |
| leak-museum | 101 | 56.44% | 60.40% | 2.29% |
| nerel | 24,369 | 56.82% | 67.23% | — |
| dialogpii-multi | 10,345 | 57.27% | 67.55% | — |
| scanpatch | 8,708 | 57.30% | 75.06% | 4.25% |
| tab-echr | 3,830 | 58.17% | 76.06% | — |
| corrupt-redmadrobot | 5,531 | 58.99% | 63.24% | 1.48% |
| jayguard | 1,195 | 61.09% | 75.40% | 1.39% |
| synth-secrets-ru | 581 | 61.62% | 63.86% | 12.01% |
| synth-secrets-en | 581 | 61.96% | 64.37% | 11.22% |
| redact-multi | 22,864 | 62.17% | 82.68% | — |
| redact-ru | 9,230 | 63.53% | 87.10% | 10.06% |
| secrets-rules | 746 | 65.55% | 66.62% | 32.61% |
| corrupt-hivetrace | 1,667 | 68.39% | 78.64% | 5.96% |
| redmadrobot | 5,516 | 72.10% | 74.13% | 0.97% |
| alrosait | 1,862 | 72.40% | 85.93% | 7.97% |
| dialogpii-en | 3,087 | 73.28% | 82.57% | 8.38% |
| factrueval | 7,966 | 73.75% | 74.83% | — |
| nym-multi | 8,169 | 77.50% | 79.56% | 8.05% |
| creddata | 776 | 78.35% | 81.96% | 6.94% |
| privy | 1,899 | 78.78% | 82.62% | 7.34% |
| synth-jira-comments | 2,705 | 81.44% | 86.36% | — |
| nym-ru | 9,497 | 82.17% | 85.86% | — |
| ameau01 | 2,242 | 82.87% | 83.72% | 5.17% |
| hivetrace | 1,667 | 84.64% | 88.00% | 7.49% |
| nym-en | 3,584 | 86.19% | 88.73% | 9.92% |
| tonicai | 2,417 | 88.75% | 92.93% | 1.45% |
| kiji-multi | 7,565 | 90.64% | 93.23% | — |
| russian-pii-66k | 4,805 | 91.09% | 96.07% | — |
| kiji-en | 7,623 | 93.11% | 96.37% | — |
| arthur-passwords | 280 | 97.86% | 97.86% | 4.12% |