PII Arena
ONNX detector

bardsai-eu PII detection benchmark

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

66.62%fully hidden annotations
81.37%detected / overlapped
7.72%extra masking
CPU characters / second

Strengths

Highest complete-masking rates by normalized data type.

People's names90.73%59,642 / 65,736
Logins & usernames80.01%9,107 / 11,382
Addresses & locations75.76%30,002 / 39,602

Watch list

Lowest observed complete-masking rates; inspect before deployment.

Network identifiers10.62%2,721 / 25,631
Other sensitive attributes39.51%1,267 / 3,207
Phone numbers & email43.27%9,511 / 21,981

Language coverage

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

LanguageDatasetsFully hidden ↑Detected ↑Extra masking ↓
English1771.87%84.39%8.72%
Russian1961.12%78.19%5.61%
Multilingual576.97%87.36%2.07%

Configuration, source and version

Family
ONNX
Revision
0e72e19f030ed4e661b1673e549af8e0dd176386
Eligible datasets
41 / 41
Character F1
0.703
Untouched diagnostic
18.63% · 42,383 / 227,466
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 ↓
rubai-ru3,45818.16%94.82%
multiconer-ru1,20822.02%30.05%3.61%
alexen21,26126.33%92.94%11.46%
leak-museum10126.73%38.61%4.46%
privy1,89935.02%48.87%6.10%
synth-ru-tickets32,09935.09%55.52%
ameau012,24244.69%54.82%3.27%
synth-jira-comments2,70544.81%60.96%
synth-wiki-tables13,09445.60%81.84%
corrupt-redmadrobot5,53153.34%64.71%1.15%
tonicai2,41753.58%69.47%0.25%
leaky-repo9553.68%78.95%2.32%
synth-env-configs39256.89%89.80%23.30%
corrupt-hivetrace1,66759.51%80.26%0.25%
redmadrobot5,51662.89%73.97%1.84%
creddata77664.56%75.77%6.15%
dialogpii-multi10,34564.74%77.06%
tab-echr3,83065.46%86.55%
alrosait1,86265.95%97.10%5.33%
gretel-multi4,48466.57%81.51%2.09%
hivetrace1,66767.07%85.54%0.31%
jayguard1,19568.95%82.85%2.20%
synth-secrets-ru58170.22%90.53%28.24%
corrupt-secrets-issues28870.83%94.44%5.73%
dialogpii-en3,08771.66%82.93%5.86%
secrets-issues28872.57%95.49%5.97%
scanpatch8,70872.58%81.73%1.26%
nym-en3,58473.47%83.04%0.39%
synth-secrets-en58173.49%92.94%27.99%
nym-ru9,49774.20%86.10%
nym-multi8,16976.80%84.43%0.94%
redact-ru9,23077.83%92.13%9.64%
nemotron-pii9,39178.42%92.39%6.68%
redact-multi22,86479.98%91.65%
russian-pii-66k4,80581.58%96.57%
secrets-rules74682.17%90.21%37.30%
nerel24,36985.55%90.18%
kiji-en7,62389.69%95.82%
kiji-multi7,56590.96%95.14%
factrueval7,96692.96%94.15%
arthur-passwords28095.00%95.00%14.53%