PII Arena
HF detector

ru-legal-ner+cpu-int8 PII detection benchmark

Measured on 9 of 41 retained datasets under the same frozen scoring and masking protocol.

36.07%fully hidden annotations
64.97%detected / overlapped
8.12%extra masking
CPU characters / second

Strengths

Highest complete-masking rates by normalized data type.

Phone numbers & email66.00%4,283 / 6,489
Documents & identifiers61.98%3,676 / 5,931
People's names58.23%6,428 / 11,039

Watch list

Lowest observed complete-masking rates; inspect before deployment.

Network identifiers2.15%407 / 18,902
Other sensitive attributes13.84%31 / 224
Dates & times20.31%378 / 1,861

Language coverage

The overall leader can differ from the best choice for one language slice.

LanguageDatasetsFully hidden ↑Detected ↑Extra masking ↓
English420.78%39.23%8.25%
Russian439.88%74.99%7.76%
Multilingual143.37%53.99%

Configuration, source and version

Family
HF
Revision
c0313a42ac147ccc38f5c3b0b3e77cab53208683
Eligible datasets
9 / 41
Character F1
0.453
Untouched diagnostic
35.03% · 25,254 / 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.

DatasetGold annotationsFully hidden ↑Detected ↑Extra masking ↓
tab-echr3,8306.08%19.92%
ameau012,24217.13%32.07%1.90%
nemotron-pii9,39126.64%47.44%21.26%
synth-ru-tickets32,09927.61%72.67%
kiji-multi7,56543.37%53.99%
secrets-issues28853.47%84.03%9.44%
redmadrobot5,51660.03%75.16%8.78%
nym-ru9,49764.17%80.40%
hivetrace1,66771.03%88.18%1.02%