PPLX detector
pplx PII detection benchmark
Measured on 41 of 41 retained datasets under the same frozen scoring and masking protocol.
74.92%fully hidden annotations
79.05%detected / overlapped
11.42%extra masking
1,006CPU characters / second
Strengths
Highest complete-masking rates by normalized data type.
Logins & usernames98.87%11,253 / 11,382
Dates & times96.26%5,296 / 5,502
Phone numbers & email94.90%20,861 / 21,981
Watch list
Lowest observed complete-masking rates; inspect before deployment.
Organizations5.09%981 / 19,274
Other sensitive attributes53.29%1,709 / 3,207
Addresses & locations55.49%21,974 / 39,602
Language coverage
The overall leader can differ from the best choice for one language slice.
| Language | Datasets | Fully hidden ↑ | Detected ↑ | Extra masking ↓ |
|---|---|---|---|---|
| English | 17 | 81.52% | 84.67% | 13.11% |
| Russian | 19 | 69.24% | 74.29% | 7.92% |
| Multilingual | 5 | 84.77% | 87.26% | 1.51% |
Configuration, source and version
- Family
- PPLX
- Revision
- f1f90a53823f5df0a1344c1e137d9fffdaab54d6
- Eligible datasets
- 41 / 41
- Character F1
- 0.716
- Untouched diagnostic
- 20.95% · 47,647 / 227,466
- Benchmark version
- 1.0.3 · 2026-09-09
- CPU seconds / 10k chars
- 9.940
- GPU seconds / 10k chars
- 2.197
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 | 10.70% | 99.54% | — |
| multiconer-ru | 1,208 | 11.67% | 19.29% | 2.38% |
| nerel | 24,369 | 12.42% | 12.67% | — |
| factrueval | 7,966 | 12.58% | 12.58% | — |
| tab-echr | 3,830 | 28.36% | 39.58% | — |
| alexen2 | 1,261 | 55.59% | 94.77% | 5.80% |
| dialogpii-multi | 10,345 | 65.15% | 70.84% | — |
| dialogpii-en | 3,087 | 71.01% | 75.90% | 5.64% |
| jayguard | 1,195 | 71.97% | 83.51% | 2.11% |
| secrets-issues | 288 | 75.00% | 95.83% | 9.88% |
| tonicai | 2,417 | 75.51% | 79.23% | 0.75% |
| corrupt-secrets-issues | 288 | 75.69% | 95.83% | 10.10% |
| corrupt-redmadrobot | 5,531 | 78.65% | 86.48% | 1.57% |
| redmadrobot | 5,516 | 78.86% | 85.95% | 1.59% |
| gretel-multi | 4,484 | 79.55% | 81.76% | 1.45% |
| ameau01 | 2,242 | 81.62% | 82.47% | 3.77% |
| leak-museum | 101 | 82.18% | 83.17% | 4.73% |
| leaky-repo | 95 | 85.26% | 92.63% | 3.32% |
| scanpatch | 8,708 | 85.54% | 87.55% | 1.29% |
| nemotron-pii | 9,391 | 86.19% | 87.76% | 6.93% |
| creddata | 776 | 86.60% | 94.20% | 7.23% |
| redact-multi | 22,864 | 86.98% | 89.39% | — |
| redact-ru | 9,230 | 87.12% | 90.46% | 9.50% |
| synth-wiki-tables | 13,094 | 88.89% | 94.39% | — |
| privy | 1,899 | 91.84% | 94.94% | 14.04% |
| secrets-rules | 746 | 93.43% | 96.11% | 36.71% |
| corrupt-hivetrace | 1,667 | 93.70% | 96.52% | 1.51% |
| nym-ru | 9,497 | 94.01% | 96.06% | — |
| nym-en | 3,584 | 94.89% | 96.15% | 2.15% |
| hivetrace | 1,667 | 95.14% | 97.30% | 0.97% |
| synth-ru-tickets | 32,099 | 95.59% | 97.36% | — |
| kiji-multi | 7,565 | 95.62% | 95.80% | — |
| kiji-en | 7,623 | 95.74% | 95.93% | — |
| nym-multi | 8,169 | 96.22% | 97.22% | 5.43% |
| alrosait | 1,862 | 96.67% | 99.46% | 4.06% |
| synth-secrets-en | 581 | 97.42% | 99.83% | 43.92% |
| synth-secrets-ru | 581 | 97.59% | 100.00% | 44.69% |
| russian-pii-66k | 4,805 | 97.94% | 99.77% | — |
| synth-env-configs | 392 | 98.21% | 100.00% | 29.25% |
| synth-jira-comments | 2,705 | 99.59% | 99.93% | — |
| arthur-passwords | 280 | 100.00% | 100.00% | 17.73% |