GLINER2 detector
gliner2-hivetrace-uni PII detection benchmark
Measured on 39 of 41 retained datasets under the same frozen scoring and masking protocol.
49.89%fully hidden annotations
55.11%detected / overlapped
1.27%extra masking
3,424CPU characters / second
Strengths
Highest complete-masking rates by normalized data type.
Phone numbers & email79.60%16,279 / 20,450
Logins & usernames78.36%8,684 / 11,082
Documents & identifiers69.63%9,816 / 14,097
Watch list
Lowest observed complete-masking rates; inspect before deployment.
Network identifiers2.15%529 / 24,593
Other sensitive attributes9.98%320 / 3,207
Passwords, keys & tokens20.36%1,278 / 6,278
Language coverage
The overall leader can differ from the best choice for one language slice.
| Language | Datasets | Fully hidden ↑ | Detected ↑ | Extra masking ↓ |
|---|---|---|---|---|
| English | 16 | 51.76% | 54.99% | 1.13% |
| Russian | 19 | 44.58% | 50.52% | 1.61% |
| Multilingual | 4 | 63.60% | 67.97% | 1.12% |
Configuration, source and version
- Family
- GLINER2
- Revision
- 5644cdb7c953463fa3a370e76617c03c2dd29f92
- Eligible datasets
- 39 / 41
- Character F1
- 0.571
- Untouched diagnostic
- 44.89% · 95,890 / 213,591
- Benchmark version
- 1.0.3 · 2026-09-09
- CPU seconds / 10k chars
- 2.921
- GPU seconds / 10k chars
- Not measured
source overlap: nemotron-pii, gretel-multi
Results by dataset
Complete masking is primary; detection shows whether any character overlapped.
| Dataset | Gold annotations | Fully hidden ↑ | Detected ↑ | Extra masking ↓ |
|---|---|---|---|---|
| synth-env-configs | 392 | 0.51% | 0.51% | 0.14% |
| synth-secrets-en | 581 | 0.86% | 0.86% | 0.09% |
| synth-secrets-ru | 581 | 1.03% | 1.03% | 0.21% |
| rubai-ru | 3,458 | 4.80% | 49.02% | — |
| corrupt-secrets-issues | 288 | 5.21% | 14.93% | 0.44% |
| secrets-issues | 288 | 7.99% | 13.89% | 0.45% |
| secrets-rules | 746 | 13.81% | 14.88% | 4.30% |
| factrueval | 7,966 | 14.01% | 14.24% | — |
| arthur-passwords | 280 | 15.00% | 15.00% | 1.49% |
| leaky-repo | 95 | 16.84% | 18.95% | 0.00% |
| leak-museum | 101 | 18.81% | 18.81% | 0.54% |
| alexen2 | 1,261 | 21.81% | 87.31% | 4.51% |
| creddata | 776 | 22.81% | 23.84% | 2.10% |
| tab-echr | 3,830 | 24.65% | 33.29% | — |
| synth-jira-comments | 2,705 | 35.53% | 36.56% | — |
| synth-ru-tickets | 32,099 | 39.77% | 43.43% | — |
| synth-wiki-tables | 13,094 | 40.84% | 41.74% | — |
| ameau01 | 2,242 | 41.44% | 41.53% | 0.82% |
| multiconer-ru | 1,208 | 43.05% | 55.96% | 7.58% |
| nerel | 24,369 | 45.71% | 49.78% | — |
| corrupt-redmadrobot | 5,531 | 45.81% | 54.44% | 1.84% |
| dialogpii-multi | 10,345 | 49.21% | 57.19% | — |
| nym-ru | 9,497 | 49.67% | 55.71% | — |
| corrupt-hivetrace | 1,667 | 50.69% | 64.73% | 4.19% |
| nym-multi | 8,169 | 53.34% | 56.29% | 1.12% |
| scanpatch | 8,708 | 54.16% | 58.68% | 1.41% |
| alrosait | 1,862 | 55.42% | 73.04% | 1.66% |
| redmadrobot | 5,516 | 56.74% | 59.95% | 1.11% |
| privy | 1,899 | 56.82% | 58.82% | 6.31% |
| jayguard | 1,195 | 58.74% | 71.46% | 1.68% |
| nym-en | 3,584 | 61.38% | 63.62% | 4.42% |
| russian-pii-66k | 4,805 | 62.60% | 67.83% | — |
| dialogpii-en | 3,087 | 64.20% | 73.53% | 5.04% |
| tonicai | 2,417 | 67.44% | 68.76% | 1.23% |
| kiji-multi | 7,565 | 67.85% | 68.86% | — |
| hivetrace | 1,667 | 68.99% | 71.93% | 0.51% |
| kiji-en | 7,623 | 71.43% | 72.44% | — |
| redact-multi | 22,864 | 72.38% | 76.73% | — |
| redact-ru | 9,230 | 72.42% | 79.14% | 6.93% |