PII Arena
GLINER2 detector

gliner2-vladlinv-ru PII detection benchmark

Measured on 19 of 41 retained datasets under the same frozen scoring and masking protocol.

42.50%fully hidden annotations
46.79%detected / overlapped
1.61%extra masking
569CPU characters / second

Strengths

Highest complete-masking rates by normalized data type.

Phone numbers & email83.90%12,062 / 14,377
People's names76.72%26,034 / 33,934
Documents & identifiers68.25%5,937 / 8,699

Watch list

Lowest observed complete-masking rates; inspect before deployment.

Network identifiers0.61%141 / 23,092
Organizations1.22%138 / 11,306
Other sensitive attributes4.40%49 / 1,113

Language coverage

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

LanguageDatasetsFully hidden ↑Detected ↑Extra masking ↓
Russian1942.50%46.79%1.61%

Configuration, source and version

Family
GLINER2
Revision
d4c05667b10d8e8db0f0626934135731d0530983
Eligible datasets
19 / 41
Character F1
0.586
Untouched diagnostic
53.21% · 72,584 / 136,419
Benchmark version
1.0.3 · 2026-09-09
CPU seconds / 10k chars
17.575
GPU seconds / 10k chars
Not measured

Russian labels

Results by dataset

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

DatasetGold annotationsFully hidden ↑Detected ↑Extra masking ↓
rubai-ru3,4586.62%77.01%
synth-ru-tickets32,09922.74%24.51%
alexen21,26126.33%100.00%0.00%
factrueval7,96631.50%31.56%
multiconer-ru1,20832.20%39.16%3.53%
synth-secrets-ru58133.22%35.28%8.46%
scanpatch8,70835.38%36.68%0.56%
nerel24,36936.01%37.53%
synth-jira-comments2,70539.96%40.63%
corrupt-redmadrobot5,53147.89%49.00%0.14%
redact-ru9,23048.43%50.78%6.02%
redmadrobot5,51655.00%55.53%0.13%
jayguard1,19564.60%74.81%0.68%
nym-ru9,49765.58%66.16%
corrupt-hivetrace1,66768.33%73.13%0.46%
synth-wiki-tables13,09469.07%73.66%
hivetrace1,66776.78%78.52%0.13%
russian-pii-66k4,80578.96%79.06%
alrosait1,86290.49%96.94%1.40%