MULTI · PII
dialogpii-multi PII detection benchmark results
A frozen evaluation slice with 500 rows, 10,345 normalized gold annotations and source-level provenance.
500input rows
10,345gold annotations
1,964,464characters
41measured detectors
Dataset provenance
- Language group
- MULTI
- Task
- PII
- License
- cc-by-4.0
- Data categories
- ADDRESS, CONTACT, ID, NET, ORG, PERSON
- Project generated
- No
- Corrupted copy
- No
Rows without annotations are not verified clean. Reported extra masking is not a human-confirmed false-positive rate.
Detector outcomes
Sorted by fully hidden annotations. Training-source overlaps are excluded.
| Detector | Family | Fully hidden ↑ | Detected / any overlap ↑ | Extra masking ↓ | Char F1 ↑ |
|---|---|---|---|---|---|
| pii-shield-onnx | ONNX | 74.31% | 82.24% | — | 0.119 |
| gliner2-fastino | GLINER2 | 71.70% | 77.90% | — | 0.538 |
| gliner-urchade | GLINER | 70.47% | 75.07% | — | 0.459 |
| gliner2-hivetrace-omni | GLINER2 | 66.34% | 73.96% | — | 0.604 |
| pplx | PPLX | 65.15% | 70.84% | — | 0.679 |
| bardsai-eu | ONNX | 64.74% | 77.06% | — | 0.718 |
| gliner-multi-v21 | GLINER | 64.06% | 70.09% | — | 0.458 |
| nuner-zero | GLINER | 62.71% | 68.20% | — | 0.443 |
| davlan-xlmr | HF | 60.63% | 69.42% | — | 0.680 |
| gliner25-fastino | GLINER2 | 59.15% | 64.02% | — | 0.560 |
| gliner-nvidia | GLINER | 57.27% | 67.55% | — | 0.479 |
| ner-ru-yqelz | HF | 56.33% | 65.28% | — | 0.644 |
| nym-base | HF | 55.49% | 72.36% | — | 0.696 |
| davlan-mbert | HF | 54.99% | 64.60% | — | 0.635 |
| gliner2-large | GLINER2 | 54.87% | 60.68% | — | 0.433 |
| gliner-pii-base | GLINER | 53.31% | 58.38% | — | 0.532 |
| nym-small | ONNX | 51.45% | 67.96% | — | 0.667 |
| gliner-pii-edge | GLINER | 51.06% | 56.89% | — | 0.537 |
| stanza-ru | STANZA | 49.77% | 54.96% | — | 0.164 |
| mmbert32k | HF | 49.71% | 68.42% | — | 0.485 |
| gliner2-hivetrace-uni | GLINER2 | 49.21% | 57.19% | — | 0.511 |
| gliner2-vladlinv | GLINER2 | 47.13% | 52.69% | — | 0.550 |
| ru-legal-ner | HF | 45.39% | 55.80% | — | 0.111 |
| opf-ru | OPF | 44.62% | 59.58% | — | 0.359 |
| openmed-multilingual | HF | 40.57% | 56.92% | — | 0.527 |
| ner-ru-gherman | HF | 40.02% | 56.25% | — | 0.520 |
| gliner-stream-pii | GLINER | 39.76% | 47.22% | — | 0.518 |
| gravitee-small | HF | 39.35% | 46.01% | — | 0.249 |
| traciora | OPF | 36.49% | 45.79% | — | 0.464 |
| ru-pii-ner | RUPII | 35.68% | 39.55% | — | 0.392 |
| openmed-nemotron | HF | 34.19% | 51.30% | — | 0.308 |
| kalyan-ettin | HF | 32.18% | 44.74% | — | 0.459 |
| openai-base | OPF | 31.43% | 35.87% | — | 0.363 |
| apararti | OPF | 30.81% | 38.23% | — | 0.414 |
| opf-kz-ru | OPF | 24.65% | 29.55% | — | 0.367 |
| opf-ru-v2 | OPF | 22.64% | 28.94% | — | 0.409 |
| spacy-ru-lg | SPACY | 18.37% | 20.06% | — | 0.123 |
| natasha | NATASHA | 8.43% | 8.98% | — | 0.096 |
| rules-ru | RULES | 2.14% | 2.81% | — | 0.100 |
| fef2-secret-ru | HF | 0.75% | 0.99% | — | 0.018 |
| spacy-alrosait | SPACY | 0.06% | 0.14% | — | 0.002 |