PII Arena
HF detector

ner-ru-yqelz PII detection benchmark

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

39.74%fully hidden annotations
49.75%detected / overlapped
9.17%extra masking
1,413CPU characters / second

Strengths

Highest complete-masking rates by normalized data type.

People's names74.27%48,819 / 65,736
Organizations65.64%12,651 / 19,274
Addresses & locations60.20%23,840 / 39,602

Watch list

Lowest observed complete-masking rates; inspect before deployment.

Customer & employee IDs0.81%65 / 7,994
Dates & times1.07%59 / 5,502
Phone numbers & email1.26%278 / 21,981

Language coverage

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

LanguageDatasetsFully hidden ↑Detected ↑Extra masking ↓
English1743.79%51.08%9.18%
Russian1936.52%47.88%9.35%
Multilingual545.12%53.59%4.15%

Configuration, source and version

Family
HF
Revision
4e03a3a079f5bfdb678eb9ee01235e25a7aef5f6
Eligible datasets
41 / 41
Character F1
0.445
Untouched diagnostic
50.25% · 114,298 / 227,466
Benchmark version
1.0.3 · 2026-09-09
CPU seconds / 10k chars
7.077
GPU seconds / 10k chars
0.177

Results by dataset

Complete masking is primary; detection shows whether any character overlapped.

DatasetGold annotationsFully hidden ↑Detected ↑Extra masking ↓
synth-env-configs3921.53%15.56%8.55%
arthur-passwords2803.93%3.93%10.17%
rubai-ru3,4584.51%56.28%
synth-secrets-ru5815.16%9.29%13.91%
synth-secrets-en5815.68%11.36%14.06%
leak-museum1015.94%7.92%8.55%
synth-ru-tickets32,09910.54%29.85%
leaky-repo9513.68%18.95%14.58%
synth-wiki-tables13,09414.42%18.24%
secrets-rules74618.77%28.02%27.26%
corrupt-hivetrace1,66719.56%33.23%5.19%
corrupt-secrets-issues28819.79%33.68%7.56%
synth-jira-comments2,70521.74%24.62%
privy1,89923.49%34.39%1.94%
secrets-issues28825.35%34.38%8.70%
alexen21,26126.09%55.99%18.61%
hivetrace1,66726.87%37.79%5.96%
nym-ru9,49729.78%38.33%
gretel-multi4,48429.88%47.26%4.21%
russian-pii-66k4,80532.47%40.60%
corrupt-redmadrobot5,53132.53%42.27%8.18%
nym-multi8,16933.10%37.24%0.00%
nym-en3,58438.48%42.24%4.79%
ameau012,24238.76%42.69%12.88%
nemotron-pii9,39140.52%46.81%14.58%
redmadrobot5,51641.53%50.53%9.74%
creddata77643.04%54.77%10.84%
scanpatch8,70844.89%52.04%4.12%
redact-multi22,86445.79%55.84%
redact-ru9,23046.01%55.62%16.05%
kiji-multi7,56549.76%52.20%
kiji-en7,62351.65%55.06%
alrosait1,86254.99%98.93%4.33%
tab-echr3,83055.46%75.25%
dialogpii-multi10,34556.33%65.28%
tonicai2,41757.34%64.63%5.25%
dialogpii-en3,08760.12%66.76%4.06%
jayguard1,19560.33%82.59%6.22%
nerel24,36969.29%74.06%
factrueval7,96678.77%79.94%
multiconer-ru1,20893.54%96.52%35.17%