ONNX detector
ner-ru-gherman-onnx PII detection benchmark
Measured on 9 of 41 retained datasets under the same frozen scoring and masking protocol.
19.10%fully hidden annotations
36.49%detected / overlapped
0.12%extra masking
—CPU characters / second
Strengths
Highest complete-masking rates by normalized data type.
People's names63.95%7,059 / 11,039
Addresses & locations54.89%4,889 / 8,907
Logins & usernames19.21%1,689 / 8,791
Watch list
Lowest observed complete-masking rates; inspect before deployment.
Other sensitive attributes0.00%0 / 224
Customer & employee IDs0.00%0 / 1,819
Network identifiers0.00%0 / 18,902
Language coverage
The overall leader can differ from the best choice for one language slice.
| Language | Datasets | Fully hidden ↑ | Detected ↑ | Extra masking ↓ |
|---|---|---|---|---|
| English | 4 | 23.92% | 41.31% | 0.10% |
| Russian | 4 | 11.98% | 31.21% | 0.17% |
| Multilingual | 1 | 54.95% | 60.49% | — |
Configuration, source and version
- Family
- ONNX
- Revision
- 2bd219a78a06657c59fa8c2540ccbad703972e7c
- Eligible datasets
- 9 / 41
- Character F1
- 0.355
- Untouched diagnostic
- 63.51% · 45,789 / 72,095
- Benchmark version
- 1.0.3 · 2026-09-09
- CPU seconds / 10k chars
- Not measured
- GPU seconds / 10k chars
- Not measured
mirror of ner-ru-gherman
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% | 0.11% |
| hivetrace | 1,667 | 0.84% | 30.17% | 0.09% |
| synth-ru-tickets | 32,099 | 4.70% | 27.12% | — |
| ameau01 | 2,242 | 10.26% | 29.30% | 0.06% |
| tab-echr | 3,830 | 14.39% | 55.95% | — |
| nym-ru | 9,497 | 25.25% | 37.61% | — |
| nemotron-pii | 9,391 | 31.80% | 39.46% | 0.00% |
| redmadrobot | 5,516 | 34.86% | 44.29% | 0.19% |
| kiji-multi | 7,565 | 54.95% | 60.49% | — |