PII Arena
GLINER2 detector

gliner2-hivetrace-omni-ru PII detection benchmark

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

71.70%fully hidden annotations
78.21%detected / overlapped
3.07%extra masking
1,163CPU characters / second

Strengths

Highest complete-masking rates by normalized data type.

Phone numbers & email91.11%13,099 / 14,377
Network identifiers81.34%18,782 / 23,092
People's names74.47%25,271 / 33,934

Watch list

Lowest observed complete-masking rates; inspect before deployment.

Other sensitive attributes43.58%485 / 1,113
Dates & times43.95%1,159 / 2,637
Addresses & locations55.78%12,449 / 22,319

Language coverage

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

LanguageDatasetsFully hidden ↑Detected ↑Extra masking ↓
Russian1971.70%78.21%3.07%

Configuration, source and version

Family
GLINER2
Revision
763f8aa9a7714dd2c3264f3751acdd5d854c5ac4
Eligible datasets
19 / 41
Character F1
0.808
Untouched diagnostic
21.79% · 29,721 / 136,419
Benchmark version
1.0.3 · 2026-09-09
CPU seconds / 10k chars
8.598
GPU seconds / 10k chars
Not measured

Russian labels; source overlap: nemotron-pii, gretel-multi

Results by dataset

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

DatasetGold annotationsFully hidden ↑Detected ↑Extra masking ↓
rubai-ru3,4587.23%72.04%
alexen21,26127.68%99.60%8.95%
synth-secrets-ru58141.65%41.82%8.13%
scanpatch8,70854.09%62.26%2.13%
corrupt-redmadrobot5,53157.51%64.53%0.88%
nerel24,36964.09%68.09%
redmadrobot5,51666.30%72.03%0.78%
factrueval7,96667.99%68.77%
redact-ru9,23070.68%78.28%13.51%
jayguard1,19571.21%84.77%1.89%
multiconer-ru1,20872.68%88.74%18.21%
corrupt-hivetrace1,66773.79%81.76%1.26%
nym-ru9,49774.77%81.06%
synth-jira-comments2,70576.75%77.41%
alrosait1,86278.84%89.53%8.04%
synth-wiki-tables13,09480.92%81.66%
hivetrace1,66784.04%87.82%0.59%
synth-ru-tickets32,09987.42%89.95%
russian-pii-66k4,80587.49%93.88%