GLINER detector
gliner-pii-base+cpu-int8 PII detection benchmark
Measured on 9 of 41 retained datasets under the same frozen scoring and masking protocol.
0.61%fully hidden annotations
0.62%detected / overlapped
<0.01%extra masking
—CPU characters / second
Strengths
Highest complete-masking rates by normalized data type.
Bank accounts & cards5.05%95 / 1,880
Organizations2.04%93 / 4,568
Documents & identifiers1.91%113 / 5,931
Watch list
Lowest observed complete-masking rates; inspect before deployment.
Other sensitive attributes0.00%0 / 224
Network identifiers0.00%0 / 18,902
Logins & usernames0.00%0 / 8,791
Language coverage
The overall leader can differ from the best choice for one language slice.
| Language | Datasets | Fully hidden ↑ | Detected ↑ | Extra masking ↓ |
|---|---|---|---|---|
| English | 4 | 2.29% | 2.32% | <0.01% |
| Russian | 4 | 0.15% | 0.16% | 0.00% |
| Multilingual | 1 | 0.04% | 0.04% | — |
Configuration, source and version
- Family
- GLINER
- Revision
- 61726e0ad791dcab3e29339bbec3ad42ded65641
- Eligible datasets
- 9 / 41
- Character F1
- 0.011
- Untouched diagnostic
- 99.38% · 71,651 / 72,095
- Benchmark version
- 1.0.3 · 2026-09-09
- CPU seconds / 10k chars
- Not measured
- GPU seconds / 10k chars
- Not measured
Results by dataset
Complete masking is primary; detection shows whether any character overlapped.
| Dataset | Gold annotations | Fully hidden ↑ | Detected ↑ | Extra masking ↓ |
|---|---|---|---|---|
| secrets-issues | 288 | 0.00% | 0.00% | 0.00% |
| synth-ru-tickets | 32,099 | 0.00% | 0.00% | — |
| nym-ru | 9,497 | 0.01% | 0.01% | — |
| kiji-multi | 7,565 | 0.04% | 0.04% | — |
| redmadrobot | 5,516 | 0.29% | 0.31% | 0.00% |
| ameau01 | 2,242 | 0.67% | 0.67% | 0.02% |
| tab-echr | 3,830 | 2.35% | 2.45% | — |
| nemotron-pii | 9,391 | 2.72% | 2.73% | 0.00% |
| hivetrace | 1,667 | 3.48% | 3.48% | 0.00% |