GLINER detector
gliner-stream-pii PII detection benchmark
Measured on 41 of 41 retained datasets under the same frozen scoring and masking protocol.
51.67%fully hidden annotations
63.82%detected / overlapped
3.55%extra masking
1,219CPU characters / second
Strengths
Highest complete-masking rates by normalized data type.
Logins & usernames81.39%9,264 / 11,382
Phone numbers & email64.13%14,096 / 21,981
Passwords, keys & tokens60.96%4,188 / 6,870
Watch list
Lowest observed complete-masking rates; inspect before deployment.
Other sensitive attributes20.52%658 / 3,207
Network identifiers42.18%10,810 / 25,631
Documents & identifiers43.27%6,505 / 15,034
Language coverage
The overall leader can differ from the best choice for one language slice.
| Language | Datasets | Fully hidden ↑ | Detected ↑ | Extra masking ↓ |
|---|---|---|---|---|
| English | 17 | 75.39% | 81.78% | 4.13% |
| Russian | 19 | 47.51% | 61.90% | 2.31% |
| Multilingual | 5 | 45.57% | 56.07% | 0.54% |
Configuration, source and version
- Family
- GLINER
- Revision
- e871777dc4b3b688747a0433fff8d94a36fcc7b0
- Eligible datasets
- 41 / 41
- Character F1
- 0.636
- Untouched diagnostic
- 36.18% · 82,302 / 227,466
- Benchmark version
- 1.0.3 · 2026-09-09
- CPU seconds / 10k chars
- 8.203
- GPU seconds / 10k chars
- 0.208
disputed scope
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 | 4.19% | 85.83% | — |
| alexen2 | 1,261 | 22.13% | 84.93% | 20.99% |
| corrupt-secrets-issues | 288 | 27.78% | 55.90% | 2.39% |
| redact-ru | 9,230 | 30.16% | 43.76% | 4.42% |
| multiconer-ru | 1,208 | 32.45% | 57.95% | 11.60% |
| corrupt-redmadrobot | 5,531 | 32.47% | 54.33% | 1.03% |
| redact-multi | 22,864 | 35.81% | 47.61% | — |
| scanpatch | 8,708 | 38.81% | 51.62% | 3.94% |
| dialogpii-multi | 10,345 | 39.76% | 47.22% | — |
| synth-secrets-ru | 581 | 40.62% | 47.50% | 4.26% |
| secrets-issues | 288 | 40.97% | 57.99% | 2.77% |
| factrueval | 7,966 | 41.14% | 42.93% | — |
| synth-secrets-en | 581 | 41.31% | 48.54% | 4.13% |
| corrupt-hivetrace | 1,667 | 41.81% | 70.85% | 1.62% |
| gretel-multi | 4,484 | 41.86% | 70.81% | 0.45% |
| leaky-repo | 95 | 42.11% | 58.95% | 1.34% |
| synth-env-configs | 392 | 42.35% | 48.21% | 3.96% |
| jayguard | 1,195 | 45.02% | 61.51% | 1.09% |
| alrosait | 1,862 | 45.11% | 77.39% | 6.24% |
| nerel | 24,369 | 49.35% | 55.46% | — |
| redmadrobot | 5,516 | 50.80% | 69.36% | 1.06% |
| nym-multi | 8,169 | 51.65% | 57.50% | 7.30% |
| dialogpii-en | 3,087 | 53.35% | 60.06% | 1.99% |
| synth-wiki-tables | 13,094 | 53.71% | 57.10% | — |
| nym-ru | 9,497 | 54.12% | 64.47% | — |
| leak-museum | 101 | 54.46% | 61.39% | 2.75% |
| synth-jira-comments | 2,705 | 54.82% | 60.48% | — |
| synth-ru-tickets | 32,099 | 55.16% | 72.61% | — |
| tonicai | 2,417 | 59.62% | 74.10% | 5.36% |
| russian-pii-66k | 4,805 | 64.60% | 78.88% | — |
| privy | 1,899 | 66.14% | 68.04% | 0.93% |
| tab-echr | 3,830 | 68.15% | 80.63% | — |
| nym-en | 3,584 | 69.14% | 74.19% | 7.96% |
| hivetrace | 1,667 | 69.41% | 86.20% | 1.15% |
| arthur-passwords | 280 | 70.71% | 70.71% | 3.44% |
| creddata | 776 | 77.84% | 85.44% | 6.89% |
| secrets-rules | 746 | 78.28% | 95.04% | 32.70% |
| kiji-multi | 7,565 | 78.69% | 83.46% | — |
| ameau01 | 2,242 | 82.25% | 84.26% | 6.72% |
| kiji-en | 7,623 | 85.82% | 90.06% | — |
| nemotron-pii | 9,391 | 90.08% | 94.12% | 1.40% |