PII Arena
OPF detector

openai-base PII detection benchmark

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

59.69%fully hidden annotations
65.39%detected / overlapped
8.83%extra masking
2,378CPU characters / second

Strengths

Highest complete-masking rates by normalized data type.

Phone numbers & email89.33%19,635 / 21,981
Dates & times87.71%4,826 / 5,502
Passwords, keys & tokens85.56%5,878 / 6,870

Watch list

Lowest observed complete-masking rates; inspect before deployment.

Organizations6.16%1,188 / 19,274
Other sensitive attributes11.44%367 / 3,207
Addresses & locations39.64%15,698 / 39,602

Language coverage

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

LanguageDatasetsFully hidden ↑Detected ↑Extra masking ↓
English1766.96%69.46%8.90%
Russian1956.63%63.95%9.00%
Multilingual562.39%66.21%1.46%

Configuration, source and version

Family
OPF
Revision
7ffa9a043d54d1be65afb281eddf0ffbe629385b
Eligible datasets
41 / 41
Character F1
0.657
Untouched diagnostic
34.61% · 78,725 / 227,466
Benchmark version
1.0.3 · 2026-09-09
CPU seconds / 10k chars
4.205
GPU seconds / 10k chars
1.804

Results by dataset

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

DatasetGold annotationsFully hidden ↑Detected ↑Extra masking ↓
multiconer-ru1,2084.22%7.12%1.33%
nerel24,3698.88%9.70%
factrueval7,96614.02%14.16%
alexen21,26120.86%87.87%8.63%
tab-echr3,83022.22%23.13%
rubai-ru3,45826.17%93.18%
dialogpii-multi10,34531.43%35.87%
jayguard1,19546.03%54.98%1.85%
corrupt-redmadrobot5,53151.82%59.36%1.79%
ameau012,24253.70%55.04%2.58%
dialogpii-en3,08754.13%57.60%12.89%
leak-museum10156.44%63.37%1.15%
leaky-repo9560.00%70.53%2.20%
synth-jira-comments2,70560.78%66.77%
privy1,89963.82%68.83%8.68%
gretel-multi4,48464.74%71.19%1.44%
redmadrobot5,51665.17%70.58%1.39%
tonicai2,41765.37%68.51%0.30%
corrupt-secrets-issues28868.06%89.58%4.41%
redact-multi22,86468.17%72.01%
corrupt-hivetrace1,66768.63%77.50%2.48%
nemotron-pii9,39168.64%70.27%2.55%
scanpatch8,70870.22%74.39%1.30%
nym-multi8,16970.94%74.88%3.18%
secrets-issues28871.18%92.71%4.09%
synth-ru-tickets32,09973.68%82.96%
alrosait1,86273.85%87.16%5.39%
kiji-multi7,56576.66%77.83%
redact-ru9,23076.94%80.66%11.76%
synth-env-configs39277.30%95.15%14.20%
nym-en3,58478.43%81.78%1.91%
synth-wiki-tables13,09480.89%87.03%
nym-ru9,49782.37%87.67%
kiji-en7,62384.36%84.81%
synth-secrets-en58184.51%92.94%49.46%
hivetrace1,66786.86%90.10%1.70%
creddata77688.40%91.11%4.64%
russian-pii-66k4,80589.82%92.20%
synth-secrets-ru58191.74%97.93%51.56%
arthur-passwords28092.86%92.86%13.82%
secrets-rules74697.99%99.06%45.69%