PII Arena
GLINER detector

gliner-multi-v21-ru PII detection benchmark

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

43.02%fully hidden annotations
52.58%detected / overlapped
1.60%extra masking
1,491CPU characters / second

Strengths

Highest complete-masking rates by normalized data type.

People's names66.71%22,637 / 33,934
Organizations59.18%6,691 / 11,306
Logins & usernames58.37%5,842 / 10,009

Watch list

Lowest observed complete-masking rates; inspect before deployment.

Customer & employee IDs8.03%324 / 4,036
Bank accounts & cards8.19%238 / 2,906
Phone numbers & email15.42%2,217 / 14,377

Language coverage

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

LanguageDatasetsFully hidden ↑Detected ↑Extra masking ↓
Russian1943.02%52.58%1.60%

Configuration, source and version

Family
GLINER
Revision
443d26d654e0324125a96bebd8e796c14ff2efe6
Eligible datasets
19 / 41
Character F1
0.553
Untouched diagnostic
47.42% · 64,690 / 136,419
Benchmark version
1.0.3 · 2026-09-09
CPU seconds / 10k chars
6.707
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-ru3,4585.64%83.05%
synth-wiki-tables13,09411.97%13.92%
synth-secrets-ru58116.18%22.03%1.50%
redact-ru9,23023.95%32.08%2.54%
alexen21,26126.72%84.06%3.48%
corrupt-redmadrobot5,53127.93%31.62%0.62%
corrupt-hivetrace1,66731.43%47.69%0.71%
redmadrobot5,51638.81%41.82%0.93%
synth-ru-tickets32,09939.08%51.80%
nym-ru9,49739.38%44.42%
synth-jira-comments2,70539.85%46.95%
hivetrace1,66741.63%55.73%1.06%
multiconer-ru1,20842.72%59.11%6.70%
scanpatch8,70845.99%50.71%2.71%
russian-pii-66k4,80560.69%71.59%
nerel24,36968.28%72.77%
jayguard1,19569.29%82.68%1.39%
alrosait1,86270.89%99.19%6.80%
factrueval7,96672.63%73.45%