GLINER2 detector
gliner2-hivetrace-omni PII detection benchmark
Measured on 39 of 41 retained datasets under the same frozen scoring and masking protocol.
74.82%fully hidden annotations
80.70%detected / overlapped
3.12%extra masking
676CPU characters / second
Strengths
Highest complete-masking rates by normalized data type.
Phone numbers & email90.17%18,439 / 20,450
Logins & usernames87.61%9,709 / 11,082
People's names81.57%49,949 / 61,233
Watch list
Lowest observed complete-masking rates; inspect before deployment.
Other sensitive attributes42.94%1,377 / 3,207
Dates & times49.97%2,601 / 5,205
Passwords, keys & tokens57.22%3,592 / 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 | 78.74% | 83.38% | 3.12% |
| Russian | 19 | 73.32% | 79.80% | 3.12% |
| Multilingual | 4 | 76.75% | 81.68% | 0.00% |
Configuration, source and version
- Family
- GLINER2
- Revision
- 763f8aa9a7714dd2c3264f3751acdd5d854c5ac4
- Eligible datasets
- 39 / 41
- Character F1
- 0.760
- Untouched diagnostic
- 19.30% · 41,214 / 213,591
- Benchmark version
- 1.0.3 · 2026-09-09
- CPU seconds / 10k chars
- 14.793
- GPU seconds / 10k chars
- 0.287
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 ↓ |
|---|---|---|---|---|
| rubai-ru | 3,458 | 7.26% | 74.58% | — |
| corrupt-secrets-issues | 288 | 16.32% | 30.21% | 1.30% |
| secrets-rules | 746 | 17.96% | 17.96% | 14.38% |
| secrets-issues | 288 | 20.83% | 35.07% | 1.27% |
| alexen2 | 1,261 | 27.68% | 99.60% | 9.01% |
| leak-museum | 101 | 43.56% | 44.55% | 2.83% |
| creddata | 776 | 49.61% | 51.93% | 2.85% |
| leaky-repo | 95 | 51.58% | 60.00% | 0.67% |
| corrupt-redmadrobot | 5,531 | 52.31% | 59.52% | 0.89% |
| synth-secrets-ru | 581 | 53.01% | 53.36% | 7.62% |
| synth-env-configs | 392 | 54.08% | 54.08% | 7.52% |
| synth-secrets-en | 581 | 54.22% | 54.56% | 6.50% |
| corrupt-hivetrace | 1,667 | 57.71% | 63.23% | 1.13% |
| scanpatch | 8,708 | 59.11% | 66.69% | 2.22% |
| redmadrobot | 5,516 | 63.51% | 68.44% | 0.61% |
| hivetrace | 1,667 | 65.87% | 68.63% | 0.42% |
| dialogpii-multi | 10,345 | 66.34% | 73.96% | — |
| nerel | 24,369 | 67.04% | 71.10% | — |
| factrueval | 7,966 | 69.81% | 70.76% | — |
| nym-multi | 8,169 | 70.19% | 73.34% | 0.00% |
| tab-echr | 3,830 | 71.44% | 89.35% | — |
| nym-ru | 9,497 | 72.82% | 79.58% | — |
| jayguard | 1,195 | 73.72% | 86.19% | 2.38% |
| multiconer-ru | 1,208 | 73.92% | 90.07% | 17.85% |
| privy | 1,899 | 73.99% | 75.57% | 5.72% |
| dialogpii-en | 3,087 | 75.25% | 83.22% | 6.73% |
| nym-en | 3,584 | 75.81% | 78.66% | 0.50% |
| redact-ru | 9,230 | 77.79% | 84.49% | 14.27% |
| redact-multi | 22,864 | 78.44% | 84.03% | — |
| alrosait | 1,862 | 78.79% | 89.37% | 9.82% |
| synth-wiki-tables | 13,094 | 79.61% | 79.90% | — |
| russian-pii-66k | 4,805 | 86.85% | 93.40% | — |
| synth-jira-comments | 2,705 | 88.65% | 89.39% | — |
| tonicai | 2,417 | 89.33% | 92.93% | 2.08% |
| synth-ru-tickets | 32,099 | 91.21% | 94.01% | — |
| kiji-multi | 7,565 | 92.97% | 94.18% | — |
| ameau01 | 2,242 | 94.60% | 94.65% | 6.48% |
| kiji-en | 7,623 | 95.04% | 95.64% | — |
| arthur-passwords | 280 | 98.57% | 98.57% | 6.79% |