GLINER detector
gliner-pii-edge PII detection benchmark
Measured on 41 of 41 retained datasets under the same frozen scoring and masking protocol.
57.10%fully hidden annotations
68.30%detected / overlapped
9.54%extra masking
1,454CPU characters / second
Strengths
Highest complete-masking rates by normalized data type.
Logins & usernames77.68%8,841 / 11,382
Phone numbers & email74.46%16,366 / 21,981
Passwords, keys & tokens72.87%5,006 / 6,870
Watch list
Lowest observed complete-masking rates; inspect before deployment.
Other sensitive attributes32.77%1,051 / 3,207
Network identifiers33.38%8,556 / 25,631
Dates & times42.91%2,361 / 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 | 69.42% | 75.19% | 10.96% |
| Russian | 19 | 56.78% | 71.80% | 6.58% |
| Multilingual | 5 | 49.23% | 54.52% | 1.47% |
Configuration, source and version
- Family
- GLINER
- Revision
- 9b7f39b0a2da971a5beea78d35f1539d4009c891
- Eligible datasets
- 41 / 41
- Character F1
- 0.629
- Untouched diagnostic
- 31.70% · 72,103 / 227,466
- Benchmark version
- 1.0.3 · 2026-09-09
- CPU seconds / 10k chars
- 6.878
- GPU seconds / 10k chars
- 5.804
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 | 6.33% | 75.53% | — |
| alexen2 | 1,261 | 21.97% | 92.86% | 18.22% |
| redact-multi | 22,864 | 41.73% | 46.82% | — |
| multiconer-ru | 1,208 | 43.38% | 71.77% | 19.92% |
| synth-ru-tickets | 32,099 | 45.22% | 66.23% | — |
| jayguard | 1,195 | 46.95% | 71.21% | 2.00% |
| privy | 1,899 | 48.55% | 52.19% | 8.92% |
| alrosait | 1,862 | 50.38% | 72.72% | 5.64% |
| corrupt-redmadrobot | 5,531 | 50.46% | 57.60% | 2.03% |
| leaky-repo | 95 | 50.53% | 77.89% | 6.20% |
| dialogpii-multi | 10,345 | 51.06% | 56.89% | — |
| corrupt-secrets-issues | 288 | 52.78% | 86.46% | 9.04% |
| kiji-multi | 7,565 | 53.60% | 54.71% | — |
| synth-wiki-tables | 13,094 | 55.23% | 77.46% | — |
| gretel-multi | 4,484 | 55.71% | 72.17% | 1.29% |
| redact-ru | 9,230 | 56.38% | 67.59% | 14.06% |
| scanpatch | 8,708 | 56.99% | 67.31% | 7.81% |
| kiji-en | 7,623 | 59.11% | 60.59% | — |
| leak-museum | 101 | 59.41% | 75.25% | 7.23% |
| redmadrobot | 5,516 | 59.43% | 64.52% | 1.45% |
| nym-multi | 8,169 | 60.29% | 63.24% | 13.86% |
| synth-env-configs | 392 | 63.01% | 68.62% | 19.45% |
| corrupt-hivetrace | 1,667 | 63.23% | 74.63% | 7.20% |
| secrets-issues | 288 | 64.93% | 87.15% | 7.94% |
| nerel | 24,369 | 64.96% | 73.79% | — |
| nym-en | 3,584 | 65.96% | 67.91% | 9.50% |
| synth-jira-comments | 2,705 | 67.06% | 73.83% | — |
| nym-ru | 9,497 | 68.50% | 75.83% | — |
| synth-secrets-en | 581 | 70.57% | 83.65% | 24.36% |
| dialogpii-en | 3,087 | 71.17% | 79.79% | 2.52% |
| synth-secrets-ru | 581 | 71.26% | 85.20% | 26.68% |
| ameau01 | 2,242 | 72.26% | 72.57% | 3.80% |
| nemotron-pii | 9,391 | 73.59% | 80.70% | 0.61% |
| russian-pii-66k | 4,805 | 73.92% | 78.79% | — |
| hivetrace | 1,667 | 77.56% | 83.98% | 2.70% |
| tab-echr | 3,830 | 77.91% | 86.16% | — |
| creddata | 776 | 80.03% | 92.14% | 11.53% |
| tonicai | 2,417 | 80.47% | 88.58% | 2.86% |
| factrueval | 7,966 | 81.53% | 84.60% | — |
| secrets-rules | 746 | 88.61% | 98.26% | 53.41% |
| arthur-passwords | 280 | 99.64% | 99.64% | 5.15% |