PII Arena
GLINER2 detector

gliner2-fastino-ru PII detection benchmark

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

77.70%fully hidden annotations
86.59%detected / overlapped
2.16%extra masking
1,123CPU characters / second

Strengths

Highest complete-masking rates by normalized data type.

Logins & usernames98.55%9,864 / 10,009
Phone numbers & email92.33%13,275 / 14,377
People's names81.32%27,594 / 33,934

Watch list

Lowest observed complete-masking rates; inspect before deployment.

Other sensitive attributes45.37%505 / 1,113
Bank accounts & cards55.40%1,610 / 2,906
Dates & times56.35%1,486 / 2,637

Language coverage

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

LanguageDatasetsFully hidden ↑Detected ↑Extra masking ↓
Russian1977.70%86.59%2.16%

Configuration, source and version

Family
GLINER2
Revision
c153999da5f4c509df4322b0c6a1baf3d2c284d7
Eligible datasets
19 / 41
Character F1
0.816
Untouched diagnostic
13.41% · 18,300 / 136,419
Benchmark version
1.0.3 · 2026-09-09
CPU seconds / 10k chars
8.905
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-ru3,4587.17%68.57%
alexen21,26127.36%99.68%4.31%
synth-secrets-ru58146.47%51.12%2.82%
scanpatch8,70861.50%69.30%3.23%
redact-ru9,23065.46%73.62%12.50%
corrupt-redmadrobot5,53170.20%75.21%1.25%
multiconer-ru1,20872.85%91.97%15.66%
jayguard1,19576.90%88.37%1.27%
nym-ru9,49777.68%82.14%
alrosait1,86278.25%86.68%5.23%
corrupt-hivetrace1,66779.66%85.78%0.27%
redmadrobot5,51680.13%83.45%1.25%
synth-ru-tickets32,09982.63%95.92%
synth-jira-comments2,70583.84%84.73%
nerel24,36984.86%88.74%
synth-wiki-tables13,09485.34%90.38%
factrueval7,96687.06%87.82%
hivetrace1,66787.94%89.68%0.12%
russian-pii-66k4,80592.07%94.84%