# kiji-multi - multi / pii (1000 rows, 7565 spans, 0 negatives)

Main metric: **missed** - gold spans no predicted character touched. Score threshold 0.5; spans without a score always count. `hidden` - gold spans every character of which is covered: touching one character counts as detection, not as hiding. `hidden` is counted over normalized predictions - a prediction is stretched to whole words before it is compared, so it is what a masker that repeats the same normalization would hide; a masker that masks the raw offsets of the model hides no more than this. Char P / R / F1 are reference only. Trivial baseline, mask everything: 0 missed, 100% hidden, char P 0.150 (85.0% of the text over-masked). `rows touched` / `chars masked` count rows the source left without annotations where a mask touched something, and the characters masked in them; that absence of annotations is not evidence that those rows hold nothing sensitive, so this is not a false-alarm rate. The exact CSVs keep the technical column names `fp_rows` and `fp_chars`.

| model | missed | missed % [95% CI] | hidden | char F1 [95% CI] | P | R | entity exact | entity overlap | typed F1 | rows touched | chars masked | dropped spans | ms/row | params | train |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| pplx+cpu-int8 | **76** | 1.0% [0.8%, 1.3%] | 98.8% | 0.754 [0.744, 0.764] | 0.610 | 0.986 | 0.379 | 0.892 | 0.703 | 0/0 | 0 | 0 | 209 | 596M |  |
| nym-base | **188** | 2.5% [2.2%, 2.9%] | 95.9% | 0.904 [0.899, 0.909] | 0.856 | 0.958 | 0.788 | 0.901 | 0.426 | 0/0 | 0 | 0 | 4 | 308M |  |
| nym-small | **204** | 2.7% [2.3%, 3.1%] | 95.6% | 0.907 [0.901, 0.912] | 0.861 | 0.958 | 0.778 | 0.903 | 0.412 | 0/0 | 0 | 0 | 932 | - |  |
| nym-base+cpu-int8 | **258** | 3.4% [3.0%, 3.9%] | 94.4% | 0.903 [0.897, 0.908] | 0.860 | 0.950 | 0.773 | 0.900 | 0.426 | 0/0 | 0 | 0 | 45 | 308M |  |
| pplx | **318** | 4.2% [3.7%, 4.7%] | 95.6% | 0.865 [0.858, 0.871] | 0.792 | 0.952 | 0.485 | 0.911 | 0.827 | 0/0 | 0 | 0 | 60 | 596M |  |
| gliner2-fastino | **319** | 4.2% [3.8%, 4.7%] | 94.7% | 0.871 [0.864, 0.877] | 0.814 | 0.936 | 0.729 | 0.894 | 0.797 | 0/0 | 0 | 0 | 6 | 307M |  |
| bardsai-eu | **368** | 4.9% [4.4%, 5.3%] | 91.0% | 0.842 [0.836, 0.848] | 0.775 | 0.921 | 0.466 | 0.890 | 0.415 | 0/0 | 0 | 0 | 326 | - |  |
| gliner2-hivetrace-omni | **440** | 5.8% [5.3%, 6.4%] | 93.0% | 0.870 [0.863, 0.877] | 0.831 | 0.912 | 0.759 | 0.889 | 0.813 | 0/0 | 0 | 0 | 6 | 307M |  |
| gliner-nvidia | **512** | 6.8% [5.9%, 7.6%] | 90.6% | 0.856 [0.848, 0.864] | 0.809 | 0.908 | 0.696 | 0.876 | 0.768 | 0/0 | 0 | 0 | 12 | 445M |  |
| mmbert32k | **976** | 12.9% [12.1%, 13.7%] | 75.6% | 0.804 [0.797, 0.811] | 0.778 | 0.831 | 0.343 | 0.846 | 0.392 | 0/0 | 0 | 0 | 5 | 308M |  |
| gliner-stream-pii | **1251** | 16.5% [15.4%, 17.6%] | 78.7% | 0.809 [0.799, 0.818] | 0.844 | 0.776 | 0.696 | 0.838 | 0.724 | 0/0 | 0 | 0 | 31 | 677M |  |
| openmed-multilingual | **1263** | 16.7% [15.8%, 17.6%] | 79.5% | 0.825 [0.817, 0.834] | 0.870 | 0.785 | 0.665 | 0.843 | 0.782 | 0/0 | 0 | 0 | 32 | 1.4B |  |
| gliner2-large | **1495** | 19.8% [18.7%, 20.9%] | 78.8% | 0.803 [0.795, 0.811] | 0.787 | 0.820 | 0.579 | 0.812 | 0.719 | 0/0 | 0 | 0 | 10 | 486M |  |
| kalyan-ettin | **1516** | 20.0% [19.0%, 21.1%] | 71.8% | 0.752 [0.743, 0.760] | 0.767 | 0.738 | 0.529 | 0.796 | 0.690 | 0/0 | 0 | 0 | 5 | 68M |  |
| openai-base-onnx | **1663** | 22.0% [20.6%, 23.3%] | 76.7% | 0.790 [0.782, 0.799] | 0.805 | 0.777 | 0.417 | 0.839 | 0.717 | 0/0 | 0 | 0 | 224 | - |  |
| openai-base | **1677** | 22.2% [20.7%, 23.5%] | 76.7% | 0.789 [0.781, 0.798] | 0.804 | 0.775 | 0.416 | 0.837 | 0.715 | 0/0 | 0 | 0 | 72 | 1.4B |  |
| apararti | **1683** | 22.2% [21.1%, 23.4%] | 76.0% | 0.808 [0.801, 0.815] | 0.826 | 0.791 | 0.422 | 0.846 | 0.727 | 0/0 | 0 | 0 | 64 | 1.4B |  |
| opf-kz-ru | **1698** | 22.4% [21.4%, 23.6%] | 76.1% | 0.797 [0.789, 0.805] | 0.804 | 0.789 | 0.399 | 0.840 | 0.718 | 0/0 | 0 | 0 | 60 | 1.4B |  |
| nuner-zero | **1857** | 24.5% [23.4%, 25.8%] | 74.0% | 0.764 [0.755, 0.774] | 0.725 | 0.808 | 0.518 | 0.782 | 0.689 | 0/0 | 0 | 0 | 8 | 449M |  |
| openmed-nemotron | **1889** | 25.0% [23.8%, 26.2%] | 69.0% | 0.680 [0.671, 0.690] | 0.635 | 0.732 | 0.465 | 0.702 | 0.615 | 0/0 | 0 | 0 | 36 | 1.4B |  |
| opf-ru | **1919** | 25.4% [24.4%, 26.4%] | 66.7% | 0.758 [0.749, 0.766] | 0.815 | 0.708 | 0.444 | 0.792 | 0.663 | 0/0 | 0 | 0 | 39 | 1.4B |  |
| gravitee-small | **1970** | 26.0% [24.7%, 27.4%] | 72.1% | 0.685 [0.675, 0.695] | 0.638 | 0.739 | 0.325 | 0.756 | 0.645 | 0/0 | 0 | 0 | 5 | 29M |  |
| gravitee-small+cpu-int8 | **2020** | 26.7% [25.4%, 28.1%] | 71.1% | 0.680 [0.670, 0.691] | 0.635 | 0.732 | 0.315 | 0.751 | 0.639 | 0/0 | 0 | 0 | 15 | 29M |  |
| gliner-multi-v21 | **2090** | 27.6% [26.5%, 28.9%] | 70.7% | 0.695 [0.683, 0.707] | 0.736 | 0.658 | 0.417 | 0.766 | 0.628 | 0/0 | 0 | 0 | 4 | 289M |  |
| gliner25-fastino | **2248** | 29.7% [28.4%, 31.1%] | 69.3% | 0.755 [0.746, 0.765] | 0.838 | 0.687 | 0.574 | 0.776 | 0.701 | 0/0 | 0 | 0 | 3 | 287M |  |
| gliner2-hivetrace-uni | **2356** | 31.1% [30.1%, 32.2%] | 67.9% | 0.757 [0.748, 0.767] | 0.830 | 0.696 | 0.579 | 0.761 | 0.565 | 0/0 | 0 | 0 | 13 | 147M |  |
| traciora | **2371** | 31.3% [29.8%, 32.7%] | 64.9% | 0.730 [0.720, 0.740] | 0.798 | 0.672 | 0.321 | 0.789 | 0.667 | 0/0 | 0 | 0 | 256 | 1.4B |  |
| gliner-urchade | **2400** | 31.7% [30.7%, 32.9%] | 67.4% | 0.776 [0.766, 0.785] | 0.790 | 0.763 | 0.532 | 0.751 | 0.685 | 0/0 | 0 | 0 | 4 | 289M |  |
| pii-shield-onnx | **2409** | 31.8% [30.3%, 33.4%] | 62.5% | 0.381 [0.371, 0.391] | 0.256 | 0.743 | 0.166 | 0.564 | 0.062 | 0/0 | 0 | 0 | 1965 | - |  |
| mmbert32k+cpu-int8 | **2469** | 32.6% [31.5%, 33.7%] | 50.9% | 0.653 [0.643, 0.663] | 0.720 | 0.597 | 0.280 | 0.700 | 0.259 | 0/0 | 0 | 0 | 47 | 308M |  |
| opf-ru-v2 | **2718** | 35.9% [34.6%, 37.5%] | 60.2% | 0.733 [0.723, 0.743] | 0.849 | 0.645 | 0.322 | 0.765 | 0.670 | 0/0 | 0 | 0 | 258 | 1.4B |  |
| ner-ru-gherman | **2965** | 39.2% [38.2%, 40.2%] | 55.3% | 0.576 [0.564, 0.587] | 0.785 | 0.455 | 0.576 | 0.692 | 0.221 | 0/0 | 0 | 0 | 4 | 177M |  |
| ner-ru-gherman-onnx | **2989** | 39.5% [38.6%, 40.6%] | 55.0% | 0.572 [0.560, 0.583] | 0.786 | 0.450 | 0.576 | 0.689 | 0.218 | 0/0 | 0 | 0 | 204 | - |  |
| davlan-xlmr | **3106** | 41.1% [40.3%, 41.9%] | 57.6% | 0.625 [0.613, 0.636] | 0.764 | 0.528 | 0.448 | 0.680 | 0.219 | 0/0 | 0 | 0 | 4 | 277M |  |
| davlan-mbert | **3249** | 42.9% [42.2%, 43.7%] | 55.7% | 0.616 [0.605, 0.627] | 0.758 | 0.519 | 0.441 | 0.665 | 0.217 | 0/0 | 0 | 0 | 4 | 177M |  |
| davlan-mbert+cpu-int8 | **3295** | 43.6% [42.8%, 44.4%] | 55.3% | 0.611 [0.599, 0.622] | 0.762 | 0.510 | 0.438 | 0.661 | 0.217 | 0/0 | 0 | 0 | 31 | 177M |  |
| ner-ru-gherman+cpu-int8 | **3300** | 43.6% [42.7%, 44.7%] | 52.1% | 0.538 [0.526, 0.549] | 0.790 | 0.408 | 0.563 | 0.660 | 0.188 | 0/0 | 0 | 0 | 34 | 177M |  |
| ru-pii-ner | **3421** | 45.2% [43.4%, 47.0%] | 53.9% | 0.662 [0.650, 0.674] | 0.796 | 0.566 | 0.349 | 0.675 | 0.245 | 0/0 | 0 | 0 | 102 | 358M |  |
| gliner-pii-edge | **3426** | 45.3% [43.8%, 46.8%] | 53.6% | 0.666 [0.654, 0.677] | 0.751 | 0.599 | 0.514 | 0.643 | 0.541 | 0/0 | 0 | 0 | 176 | 45M |  |
| davlan-xlmr+cpu-int8 | **3464** | 45.8% [44.9%, 46.7%] | 51.4% | 0.592 [0.580, 0.603] | 0.772 | 0.480 | 0.422 | 0.644 | 0.215 | 0/0 | 0 | 0 | 34 | 277M |  |
| ru-legal-ner+cpu-int8 | **3481** | 46.0% [44.7%, 47.3%] | 43.4% | 0.253 [0.247, 0.261] | 0.167 | 0.528 | 0.105 | 0.426 | 0.089 | 0/0 | 0 | 0 | 9 | 29M |  |
| ru-legal-ner | **3482** | 46.0% [44.7%, 47.4%] | 43.3% | 0.253 [0.246, 0.260] | 0.167 | 0.526 | 0.106 | 0.427 | 0.089 | 0/0 | 0 | 0 | 5 | 29M |  |
| gliner2-vladlinv | **3556** | 47.0% [45.2%, 48.7%] | 52.6% | 0.654 [0.640, 0.667] | 0.824 | 0.542 | 0.406 | 0.652 | 0.492 | 0/0 | 0 | 0 | 4 | 287M |  |
| ner-ru-yqelz | **3616** | 47.8% [46.8%, 48.8%] | 49.8% | 0.563 [0.552, 0.574] | 0.741 | 0.454 | 0.375 | 0.643 | 0.174 | 0/0 | 0 | 0 | 4 | 559M |  |
| gliner-pii-base | **3782** | 50.0% [48.7%, 51.4%] | 49.2% | 0.664 [0.651, 0.676] | 0.825 | 0.555 | 0.512 | 0.620 | 0.576 | 0/0 | 0 | 0 | 4 | 166M |  |
| kalyan-ettin+cpu-int8 | **3904** | 51.6% [50.2%, 53.0%] | 41.6% | 0.528 [0.515, 0.541] | 0.737 | 0.411 | 0.299 | 0.570 | 0.499 | 0/0 | 0 | 0 | 108 | 68M |  |
| stanza-ru | **4060** | 53.7% [52.6%, 54.7%] | 44.5% | 0.302 [0.293, 0.311] | 0.247 | 0.387 | 0.142 | 0.519 | 0.014 | 0/0 | 0 | 0 | 42 | - |  |
| spacy-ru-lg | **6126** | 81.0% [79.8%, 82.1%] | 18.9% | 0.193 [0.182, 0.204] | 0.233 | 0.165 | 0.077 | 0.292 | 0.009 | 0/0 | 0 | 0 | 6 | - |  |
| gliner-pii-edge+cpu-int8 | **6145** | 81.2% [80.2%, 82.3%] | 17.5% | 0.294 [0.280, 0.309] | 0.701 | 0.186 | 0.229 | 0.301 | 0.214 | 0/0 | 0 | 0 | 535 | 45M |  |
| natasha | **6837** | 90.4% [89.6%, 91.2%] | 9.6% | 0.135 [0.125, 0.146] | 0.565 | 0.077 | 0.077 | 0.171 | 0.019 | 0/0 | 0 | 0 | 2 | - |  |
| rules-ru | **7054** | 93.2% [92.7%, 93.8%] | 6.6% | 0.269 [0.248, 0.288] | 0.965 | 0.156 | 0.097 | 0.126 | 0.264 | 0/0 | 0 | 0 | 1 | - |  |
| fef2-secret-ru | **7135** | 94.3% [93.6%, 94.9%] | 4.7% | 0.136 [0.121, 0.152] | 0.901 | 0.074 | 0.050 | 0.107 | 0.056 | 0/0 | 0 | 0 | 6 | 177M |  |
| fef2-secret-ru+cpu-int8 | **7208** | 95.3% [94.7%, 95.8%] | 3.6% | 0.103 [0.089, 0.118] | 0.933 | 0.054 | 0.032 | 0.090 | 0.048 | 0/0 | 0 | 0 | 112 | 177M |  |
| spacy-alrosait | **7557** | 99.9% [99.8%, 100.0%] | 0.1% | 0.001 [0.000, 0.003] | 0.658 | 0.001 | 0.000 | 0.002 | 0.001 | 0/0 | 0 | 0 | 6 | - |  |
| gliner-pii-base+cpu-int8 | **7562** | 100.0% [99.9%, 100.0%] | 0.0% | 0.001 [0.000, 0.003] | 0.927 | 0.001 | 0.001 | 0.001 | 0.001 | 0/0 | 0 | 0 | 186 | 166M |  |
| gliner-multi-v21+cpu-int8 | **7565** | 100.0% [100.0%, 100.0%] | 0.0% | 0.000 [0.000, 0.000] | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0/0 | 0 | 0 | 325 | 289M |  |
| gliner-nvidia+cpu-int8 | **7565** | 100.0% [100.0%, 100.0%] | 0.0% | 0.000 [0.000, 0.000] | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0/0 | 0 | 0 | 527 | 445M |  |
| gliner-urchade+cpu-int8 | **7565** | 100.0% [100.0%, 100.0%] | 0.0% | 0.000 [0.000, 0.000] | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0/0 | 0 | 0 | 225 | 289M |  |

Neighbours by rank a paired bootstrap cannot tell apart (95%, missed %) - pairs, not a transitive chain, so `a ≈ b` and `b ≈ c` do not make `a ≈ c`: nym-base ≈ nym-small; pplx ≈ gliner2-fastino; gliner-stream-pii ≈ openmed-multilingual; gliner2-large ≈ kalyan-ettin; openai-base ≈ apararti; apararti ≈ opf-kz-ru; nuner-zero ≈ openmed-nemotron; openmed-nemotron ≈ opf-ru; opf-ru ≈ gravitee-small; gravitee-small+cpu-int8 ≈ gliner-multi-v21; gliner25-fastino ≈ gliner2-hivetrace-uni; gliner2-hivetrace-uni ≈ traciora; traciora ≈ gliner-urchade; gliner-urchade ≈ pii-shield-onnx; pii-shield-onnx ≈ mmbert32k+cpu-int8; ner-ru-gherman ≈ ner-ru-gherman-onnx; davlan-mbert+cpu-int8 ≈ ner-ru-gherman+cpu-int8; ner-ru-gherman+cpu-int8 ≈ ru-pii-ner; ru-pii-ner ≈ gliner-pii-edge; gliner-pii-edge ≈ davlan-xlmr+cpu-int8; davlan-xlmr+cpu-int8 ≈ ru-legal-ner+cpu-int8; ru-legal-ner+cpu-int8 ≈ ru-legal-ner; ru-legal-ner ≈ gliner2-vladlinv; gliner2-vladlinv ≈ ner-ru-yqelz; gliner-pii-base ≈ kalyan-ettin+cpu-int8; spacy-ru-lg ≈ gliner-pii-edge+cpu-int8; spacy-alrosait ≈ gliner-pii-base+cpu-int8; gliner-pii-base+cpu-int8 ≈ gliner-multi-v21+cpu-int8; gliner-multi-v21+cpu-int8 ≈ gliner-nvidia+cpu-int8; gliner-nvidia+cpu-int8 ≈ gliner-urchade+cpu-int8

## Missed by group

| model | PERSON | ADDRESS | CONTACT | ID | NET | ACCOUNT | SECRET | ORG | DATE |
|---|---|---|---|---|---|---|---|---|---|
| spans in gold | 1866 | 3612 | 259 | 1092 | 133 | 55 | 255 | 149 | 144 |
| pplx+cpu-int8 | 10 (0.5%) | 15 (0.4%) | 0 (0.0%) | 2 (0.2%) | 2 (1.5%) | 0 (0.0%) | 8 (3.1%) | 39 (26.2%) | 0 (0.0%) |
| nym-base | 26 (1.4%) | 46 (1.3%) | 0 (0.0%) | 56 (5.1%) | 12 (9.0%) | 0 (0.0%) | 27 (10.6%) | 8 (5.4%) | 13 (9.0%) |
| nym-small | 34 (1.8%) | 58 (1.6%) | 0 (0.0%) | 43 (3.9%) | 8 (6.0%) | 1 (1.8%) | 41 (16.1%) | 8 (5.4%) | 11 (7.6%) |
| nym-base+cpu-int8 | 50 (2.7%) | 90 (2.5%) | 0 (0.0%) | 50 (4.6%) | 8 (6.0%) | 0 (0.0%) | 36 (14.1%) | 11 (7.4%) | 13 (9.0%) |
| pplx | 27 (1.4%) | 131 (3.6%) | 0 (0.0%) | 5 (0.5%) | 6 (4.5%) | 0 (0.0%) | 9 (3.5%) | 140 (94.0%) | 0 (0.0%) |
| gliner2-fastino | 79 (4.2%) | 72 (2.0%) | 0 (0.0%) | 61 (5.6%) | 80 (60.2%) | 0 (0.0%) | 12 (4.7%) | 2 (1.3%) | 13 (9.0%) |
| bardsai-eu | 21 (1.1%) | 174 (4.8%) | 0 (0.0%) | 57 (5.2%) | 22 (16.5%) | 5 (9.1%) | 65 (25.5%) | 1 (0.7%) | 23 (16.0%) |
| gliner2-hivetrace-omni | 54 (2.9%) | 139 (3.8%) | 0 (0.0%) | 121 (11.1%) | 87 (65.4%) | 1 (1.8%) | 10 (3.9%) | 14 (9.4%) | 14 (9.7%) |
| gliner-nvidia | 37 (2.0%) | 288 (8.0%) | 4 (1.5%) | 139 (12.7%) | 7 (5.3%) | 2 (3.6%) | 16 (6.3%) | 7 (4.7%) | 12 (8.3%) |
| mmbert32k | 178 (9.5%) | 644 (17.8%) | 0 (0.0%) | 41 (3.8%) | 3 (2.3%) | 18 (32.7%) | 63 (24.7%) | 29 (19.5%) | 0 (0.0%) |
| gliner-stream-pii | 359 (19.2%) | 484 (13.4%) | 33 (12.7%) | 170 (15.6%) | 83 (62.4%) | 9 (16.4%) | 45 (17.6%) | 30 (20.1%) | 38 (26.4%) |
| openmed-multilingual | 163 (8.7%) | 595 (16.5%) | 1 (0.4%) | 277 (25.4%) | 55 (41.4%) | 4 (7.3%) | 44 (17.3%) | 117 (78.5%) | 7 (4.9%) |
| gliner2-large | 437 (23.4%) | 781 (21.6%) | 0 (0.0%) | 120 (11.0%) | 90 (67.7%) | 7 (12.7%) | 35 (13.7%) | 4 (2.7%) | 21 (14.6%) |
| kalyan-ettin | 236 (12.6%) | 468 (13.0%) | 10 (3.9%) | 618 (56.6%) | 40 (30.1%) | 18 (32.7%) | 82 (32.2%) | 42 (28.2%) | 2 (1.4%) |
| openai-base-onnx | 220 (11.8%) | 1148 (31.8%) | 1 (0.4%) | 35 (3.2%) | 78 (58.6%) | 9 (16.4%) | 39 (15.3%) | 131 (87.9%) | 2 (1.4%) |
| openai-base | 224 (12.0%) | 1155 (32.0%) | 1 (0.4%) | 37 (3.4%) | 79 (59.4%) | 9 (16.4%) | 39 (15.3%) | 131 (87.9%) | 2 (1.4%) |
| apararti | 432 (23.2%) | 1048 (29.0%) | 0 (0.0%) | 11 (1.0%) | 34 (25.6%) | 12 (21.8%) | 13 (5.1%) | 133 (89.3%) | 0 (0.0%) |
| opf-kz-ru | 432 (23.2%) | 1050 (29.1%) | 0 (0.0%) | 13 (1.2%) | 28 (21.1%) | 16 (29.1%) | 19 (7.5%) | 139 (93.3%) | 1 (0.7%) |
| nuner-zero | 1261 (67.6%) | 324 (9.0%) | 1 (0.4%) | 247 (22.6%) | 0 (0.0%) | 1 (1.8%) | 12 (4.7%) | 6 (4.0%) | 5 (3.5%) |
| openmed-nemotron | 235 (12.6%) | 881 (24.4%) | 6 (2.3%) | 625 (57.2%) | 17 (12.8%) | 13 (23.6%) | 60 (23.5%) | 43 (28.9%) | 9 (6.2%) |
| opf-ru | 251 (13.5%) | 1355 (37.5%) | 1 (0.4%) | 97 (8.9%) | 60 (45.1%) | 4 (7.3%) | 21 (8.2%) | 115 (77.2%) | 15 (10.4%) |
| gravitee-small | 166 (8.9%) | 1229 (34.0%) | 5 (1.9%) | 340 (31.1%) | 27 (20.3%) | 42 (76.4%) | 135 (52.9%) | 24 (16.1%) | 2 (1.4%) |
| gravitee-small+cpu-int8 | 187 (10.0%) | 1248 (34.6%) | 6 (2.3%) | 341 (31.2%) | 30 (22.6%) | 39 (70.9%) | 139 (54.5%) | 26 (17.4%) | 4 (2.8%) |
| gliner-multi-v21 | 227 (12.2%) | 684 (18.9%) | 94 (36.3%) | 782 (71.6%) | 113 (85.0%) | 30 (54.5%) | 104 (40.8%) | 30 (20.1%) | 26 (18.1%) |
| gliner25-fastino | 859 (46.0%) | 379 (10.5%) | 1 (0.4%) | 668 (61.2%) | 120 (90.2%) | 50 (90.9%) | 113 (44.3%) | 35 (23.5%) | 23 (16.0%) |
| gliner2-hivetrace-uni | 1200 (64.3%) | 728 (20.2%) | 0 (0.0%) | 193 (17.7%) | 131 (98.5%) | 2 (3.6%) | 61 (23.9%) | 24 (16.1%) | 17 (11.8%) |
| traciora | 437 (23.4%) | 1224 (33.9%) | 2 (0.8%) | 283 (25.9%) | 115 (86.5%) | 19 (34.5%) | 47 (18.4%) | 118 (79.2%) | 126 (87.5%) |
| gliner-urchade | 1687 (90.4%) | 570 (15.8%) | 2 (0.8%) | 97 (8.9%) | 6 (4.5%) | 2 (3.6%) | 13 (5.1%) | 5 (3.4%) | 18 (12.5%) |
| pii-shield-onnx | 977 (52.4%) | 1139 (31.5%) | 4 (1.5%) | 152 (13.9%) | 4 (3.0%) | 17 (30.9%) | 48 (18.8%) | 30 (20.1%) | 38 (26.4%) |
| mmbert32k+cpu-int8 | 909 (48.7%) | 1195 (33.1%) | 49 (18.9%) | 58 (5.3%) | 38 (28.6%) | 38 (69.1%) | 85 (33.3%) | 79 (53.0%) | 18 (12.5%) |
| opf-ru-v2 | 594 (31.8%) | 1597 (44.2%) | 3 (1.2%) | 91 (8.3%) | 123 (92.5%) | 21 (38.2%) | 20 (7.8%) | 127 (85.2%) | 142 (98.6%) |
| ner-ru-gherman | 68 (3.6%) | 926 (25.6%) | 196 (75.7%) | 1091 (99.9%) | 123 (92.5%) | 40 (72.7%) | 242 (94.9%) | 135 (90.6%) | 144 (100.0%) |
| ner-ru-gherman-onnx | 71 (3.8%) | 947 (26.2%) | 193 (74.5%) | 1091 (99.9%) | 123 (92.5%) | 42 (76.4%) | 243 (95.3%) | 135 (90.6%) | 144 (100.0%) |
| davlan-xlmr | 21 (1.1%) | 1209 (33.5%) | 244 (94.2%) | 1090 (99.8%) | 106 (79.7%) | 38 (69.1%) | 253 (99.2%) | 1 (0.7%) | 144 (100.0%) |
| davlan-mbert | 44 (2.4%) | 1317 (36.5%) | 256 (98.8%) | 1089 (99.7%) | 97 (72.9%) | 48 (87.3%) | 252 (98.8%) | 2 (1.3%) | 144 (100.0%) |
| davlan-mbert+cpu-int8 | 63 (3.4%) | 1308 (36.2%) | 258 (99.6%) | 1091 (99.9%) | 120 (90.2%) | 54 (98.2%) | 253 (99.2%) | 4 (2.7%) | 144 (100.0%) |
| ner-ru-gherman+cpu-int8 | 78 (4.2%) | 1217 (33.7%) | 219 (84.6%) | 1091 (99.9%) | 123 (92.5%) | 46 (83.6%) | 244 (95.7%) | 138 (92.6%) | 144 (100.0%) |
| ru-pii-ner | 405 (21.7%) | 2543 (70.4%) | 7 (2.7%) | 99 (9.1%) | 110 (82.7%) | 16 (29.1%) | 102 (40.0%) | 135 (90.6%) | 4 (2.8%) |
| gliner-pii-edge | 899 (48.2%) | 2150 (59.5%) | 8 (3.1%) | 204 (18.7%) | 10 (7.5%) | 2 (3.6%) | 40 (15.7%) | 6 (4.0%) | 107 (74.3%) |
| davlan-xlmr+cpu-int8 | 111 (5.9%) | 1448 (40.1%) | 246 (95.0%) | 1087 (99.5%) | 122 (91.7%) | 44 (80.0%) | 250 (98.0%) | 12 (8.1%) | 144 (100.0%) |
| ru-legal-ner+cpu-int8 | 1211 (64.9%) | 1704 (47.2%) | 14 (5.4%) | 297 (27.2%) | 36 (27.1%) | 10 (18.2%) | 95 (37.3%) | 87 (58.4%) | 27 (18.8%) |
| ru-legal-ner | 1234 (66.1%) | 1706 (47.2%) | 16 (6.2%) | 271 (24.8%) | 38 (28.6%) | 11 (20.0%) | 87 (34.1%) | 94 (63.1%) | 25 (17.4%) |
| gliner2-vladlinv | 144 (7.7%) | 2768 (76.6%) | 3 (1.2%) | 317 (29.0%) | 129 (97.0%) | 3 (5.5%) | 42 (16.5%) | 148 (99.3%) | 2 (1.4%) |
| ner-ru-yqelz | 396 (21.2%) | 1456 (40.3%) | 242 (93.4%) | 918 (84.1%) | 115 (86.5%) | 54 (98.2%) | 230 (90.2%) | 62 (41.6%) | 143 (99.3%) |
| gliner-pii-base | 1667 (89.3%) | 1425 (39.5%) | 8 (3.1%) | 387 (35.4%) | 62 (46.6%) | 7 (12.7%) | 107 (42.0%) | 26 (17.4%) | 93 (64.6%) |
| kalyan-ettin+cpu-int8 | 816 (43.7%) | 1541 (42.7%) | 78 (30.1%) | 996 (91.2%) | 119 (89.5%) | 41 (74.5%) | 213 (83.5%) | 80 (53.7%) | 20 (13.9%) |
| stanza-ru | 240 (12.9%) | 2036 (56.4%) | 252 (97.3%) | 958 (87.7%) | 131 (98.5%) | 50 (90.9%) | 206 (80.8%) | 46 (30.9%) | 141 (97.9%) |
| spacy-ru-lg | 1182 (63.3%) | 2977 (82.4%) | 253 (97.7%) | 1049 (96.1%) | 128 (96.2%) | 51 (92.7%) | 234 (91.8%) | 116 (77.9%) | 136 (94.4%) |
| gliner-pii-edge+cpu-int8 | 1645 (88.2%) | 3035 (84.0%) | 196 (75.7%) | 658 (60.3%) | 115 (86.5%) | 37 (67.3%) | 223 (87.5%) | 121 (81.2%) | 115 (79.9%) |
| natasha | 1474 (79.0%) | 3285 (90.9%) | 259 (100.0%) | 1091 (99.9%) | 133 (100.0%) | 54 (98.2%) | 254 (99.6%) | 143 (96.0%) | 144 (100.0%) |
| rules-ru | 1857 (99.5%) | 3607 (99.9%) | 30 (11.6%) | 1025 (93.9%) | 1 (0.8%) | 49 (89.1%) | 215 (84.3%) | 126 (84.6%) | 144 (100.0%) |
| fef2-secret-ru | 1712 (91.7%) | 3541 (98.0%) | 160 (61.8%) | 1069 (97.9%) | 131 (98.5%) | 50 (90.9%) | 186 (72.9%) | 142 (95.3%) | 144 (100.0%) |
| fef2-secret-ru+cpu-int8 | 1696 (90.9%) | 3575 (99.0%) | 179 (69.1%) | 1083 (99.2%) | 133 (100.0%) | 52 (94.5%) | 200 (78.4%) | 146 (98.0%) | 144 (100.0%) |
| spacy-alrosait | 1866 (100.0%) | 3605 (99.8%) | 258 (99.6%) | 1092 (100.0%) | 133 (100.0%) | 55 (100.0%) | 255 (100.0%) | 149 (100.0%) | 144 (100.0%) |
| gliner-pii-base+cpu-int8 | 1866 (100.0%) | 3612 (100.0%) | 258 (99.6%) | 1090 (99.8%) | 133 (100.0%) | 55 (100.0%) | 255 (100.0%) | 149 (100.0%) | 144 (100.0%) |
| gliner-multi-v21+cpu-int8 | 1866 (100.0%) | 3612 (100.0%) | 259 (100.0%) | 1092 (100.0%) | 133 (100.0%) | 55 (100.0%) | 255 (100.0%) | 149 (100.0%) | 144 (100.0%) |
| gliner-nvidia+cpu-int8 | 1866 (100.0%) | 3612 (100.0%) | 259 (100.0%) | 1092 (100.0%) | 133 (100.0%) | 55 (100.0%) | 255 (100.0%) | 149 (100.0%) | 144 (100.0%) |
| gliner-urchade+cpu-int8 | 1866 (100.0%) | 3612 (100.0%) | 259 (100.0%) | 1092 (100.0%) | 133 (100.0%) | 55 (100.0%) | 255 (100.0%) | 149 (100.0%) | 144 (100.0%) |

## Char recall by gold type

| type | group | pplx+cpu-int8 | nym-base | nym-small | nym-base+cpu-int8 | pplx | gliner2-fastino | bardsai-eu | gliner2-hivetrace-omni | gliner-nvidia | mmbert32k | gliner-stream-pii | openmed-multilingual | gliner2-large | kalyan-ettin | openai-base-onnx | openai-base | apararti | opf-kz-ru | nuner-zero | openmed-nemotron | opf-ru | gravitee-small | gravitee-small+cpu-int8 | gliner-multi-v21 | gliner25-fastino | gliner2-hivetrace-uni | traciora | gliner-urchade | pii-shield-onnx | mmbert32k+cpu-int8 | opf-ru-v2 | ner-ru-gherman | ner-ru-gherman-onnx | davlan-xlmr | davlan-mbert | davlan-mbert+cpu-int8 | ner-ru-gherman+cpu-int8 | ru-pii-ner | gliner-pii-edge | davlan-xlmr+cpu-int8 | ru-legal-ner+cpu-int8 | ru-legal-ner | gliner2-vladlinv | ner-ru-yqelz | gliner-pii-base | kalyan-ettin+cpu-int8 | stanza-ru | spacy-ru-lg | gliner-pii-edge+cpu-int8 | natasha | rules-ru | fef2-secret-ru | fef2-secret-ru+cpu-int8 | spacy-alrosait | gliner-pii-base+cpu-int8 | gliner-multi-v21+cpu-int8 | gliner-nvidia+cpu-int8 | gliner-urchade+cpu-int8 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| BUILDINGNUM | ADDRESS | 0.998 | 0.981 | 0.983 | 0.981 | 0.994 | 0.939 | 0.932 | 0.946 | 0.878 | 0.984 | 0.793 | 0.853 | 0.748 | 0.749 | 0.860 | 0.860 | 0.892 | 0.912 | 0.882 | 0.641 | 0.820 | 0.558 | 0.552 | 0.654 | 0.832 | 0.837 | 0.772 | 0.807 | 0.705 | 0.896 | 0.738 | 0.623 | 0.621 | 0.172 | 0.082 | 0.094 | 0.554 | 0.448 | 0.367 | 0.021 | 0.530 | 0.540 | 0.375 | 0.044 | 0.512 | 0.493 | 0.148 | 0.080 | 0.123 | 0.004 | 0.017 | 0.004 | 0.004 | 0.003 | 0.000 | 0.000 | 0.000 | 0.000 |
| CITY | ADDRESS | 0.995 | 0.979 | 0.975 | 0.965 | 0.946 | 0.979 | 0.997 | 0.923 | 0.939 | 0.627 | 0.901 | 0.839 | 0.870 | 0.896 | 0.586 | 0.581 | 0.588 | 0.589 | 0.925 | 0.686 | 0.488 | 0.683 | 0.671 | 0.925 | 0.943 | 0.923 | 0.627 | 0.871 | 0.708 | 0.400 | 0.438 | 0.938 | 0.938 | 0.994 | 0.992 | 0.992 | 0.931 | 0.246 | 0.554 | 0.962 | 0.495 | 0.490 | 0.192 | 0.995 | 0.722 | 0.582 | 0.672 | 0.182 | 0.209 | 0.100 | 0.000 | 0.034 | 0.014 | 0.001 | 0.000 | 0.000 | 0.000 | 0.000 |
| COMPANYNAME | ORG | 0.697 | 0.943 | 0.946 | 0.929 | 0.041 | 0.994 | 0.998 | 0.919 | 0.942 | 0.691 | 0.787 | 0.152 | 0.976 | 0.665 | 0.087 | 0.087 | 0.082 | 0.044 | 0.952 | 0.692 | 0.167 | 0.847 | 0.834 | 0.816 | 0.775 | 0.826 | 0.176 | 0.961 | 0.748 | 0.366 | 0.154 | 0.035 | 0.035 | 0.994 | 0.988 | 0.968 | 0.026 | 0.081 | 0.956 | 0.892 | 0.369 | 0.325 | 0.003 | 0.566 | 0.805 | 0.400 | 0.696 | 0.225 | 0.197 | 0.039 | 0.188 | 0.037 | 0.012 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
| COUNTRY | ADDRESS | 0.979 | 0.975 | 0.973 | 0.933 | 0.819 | 0.992 | 0.987 | 0.989 | 0.962 | 0.292 | 0.918 | 0.406 | 0.963 | 0.911 | 0.348 | 0.345 | 0.342 | 0.313 | 0.947 | 0.807 | 0.159 | 0.741 | 0.743 | 0.951 | 0.951 | 0.940 | 0.337 | 0.955 | 0.660 | 0.290 | 0.235 | 0.953 | 0.953 | 0.990 | 0.970 | 0.972 | 0.948 | 0.352 | 0.545 | 0.957 | 0.543 | 0.514 | 0.069 | 0.979 | 0.899 | 0.424 | 0.876 | 0.447 | 0.366 | 0.377 | 0.000 | 0.056 | 0.037 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
| CREDITCARDNUMBER | ID | 0.987 | 0.953 | 0.974 | 0.951 | 0.984 | 1.000 | 0.728 | 1.000 | 0.945 | 0.915 | 0.291 | 0.946 | 1.000 | 0.883 | 0.978 | 0.979 | 0.973 | 0.984 | 0.928 | 0.897 | 0.918 | 0.968 | 0.960 | 0.293 | 0.322 | 0.152 | 0.952 | 0.984 | 0.867 | 0.787 | 0.965 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.977 | 0.987 | 0.000 | 0.459 | 0.503 | 0.858 | 0.631 | 0.843 | 0.096 | 0.013 | 0.006 | 0.118 | 0.000 | 0.332 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
| DATEOFBIRTH | DATE | 1.000 | 0.895 | 0.913 | 0.891 | 1.000 | 0.897 | 0.830 | 0.892 | 0.894 | 0.860 | 0.712 | 0.891 | 0.843 | 0.906 | 0.983 | 0.983 | 0.994 | 0.989 | 0.915 | 0.892 | 0.795 | 0.974 | 0.967 | 0.797 | 0.826 | 0.869 | 0.061 | 0.841 | 0.502 | 0.488 | 0.004 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.982 | 0.239 | 0.000 | 0.600 | 0.620 | 0.991 | 0.002 | 0.341 | 0.574 | 0.013 | 0.034 | 0.183 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
| DOMAIN | NET | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | 0.750 | 1.000 | 0.000 | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | 0.000 | 1.000 | 1.000 | 1.000 | 0.000 | 1.000 | 0.000 | 1.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 1.000 | 1.000 | 0.000 | 0.250 | 0.000 | 1.000 | 0.000 | 1.000 | 1.000 | 0.000 | 0.000 | 0.625 | 0.000 | 1.000 | 1.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
| DRIVERLICENSENUM | ID | 1.000 | 0.927 | 0.922 | 0.903 | 1.000 | 0.986 | 0.957 | 0.859 | 0.777 | 0.893 | 0.733 | 0.827 | 0.931 | 0.192 | 0.993 | 0.993 | 1.000 | 0.991 | 0.609 | 0.201 | 0.864 | 0.680 | 0.687 | 0.130 | 0.305 | 0.983 | 0.664 | 0.947 | 0.876 | 0.850 | 0.849 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.967 | 0.895 | 0.008 | 0.523 | 0.610 | 0.508 | 0.073 | 0.459 | 0.009 | 0.095 | 0.066 | 0.532 | 0.000 | 0.015 | 0.063 | 0.018 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
| EMAIL | CONTACT | 1.000 | 0.958 | 0.986 | 0.955 | 1.000 | 1.000 | 0.859 | 1.000 | 0.561 | 0.818 | 0.946 | 0.991 | 1.000 | 1.000 | 0.993 | 0.993 | 1.000 | 1.000 | 0.968 | 0.993 | 0.993 | 0.999 | 0.999 | 0.169 | 0.992 | 1.000 | 1.000 | 0.947 | 0.979 | 0.207 | 0.986 | 0.195 | 0.195 | 0.050 | 0.017 | 0.006 | 0.118 | 0.963 | 0.937 | 0.035 | 0.724 | 0.709 | 0.980 | 0.015 | 0.956 | 0.813 | 0.052 | 0.037 | 0.189 | 0.000 | 1.000 | 0.652 | 0.426 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
| FIRSTNAME | PERSON | 1.000 | 0.988 | 0.984 | 0.975 | 0.995 | 0.969 | 0.992 | 0.976 | 0.990 | 0.907 | 0.780 | 0.921 | 0.765 | 0.898 | 0.886 | 0.884 | 0.771 | 0.768 | 0.394 | 0.867 | 0.856 | 0.912 | 0.893 | 0.914 | 0.679 | 0.499 | 0.739 | 0.090 | 0.387 | 0.446 | 0.640 | 0.979 | 0.980 | 0.994 | 0.977 | 0.966 | 0.978 | 0.794 | 0.612 | 0.946 | 0.324 | 0.323 | 0.951 | 0.837 | 0.139 | 0.606 | 0.849 | 0.396 | 0.134 | 0.205 | 0.007 | 0.074 | 0.081 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
| IBAN | ID | 1.000 | 0.994 | 0.995 | 0.995 | 0.997 | 0.995 | 0.918 | 0.995 | 0.938 | 0.886 | 0.644 | 0.581 | 0.853 | 0.375 | 0.994 | 0.995 | 0.992 | 0.996 | 0.926 | 0.546 | 0.881 | 0.978 | 0.974 | 0.751 | 0.931 | 0.365 | 0.981 | 0.903 | 0.920 | 0.784 | 0.957 | 0.000 | 0.000 | 0.003 | 0.007 | 0.000 | 0.000 | 0.932 | 0.693 | 0.001 | 0.679 | 0.681 | 0.940 | 0.441 | 0.755 | 0.115 | 0.086 | 0.023 | 0.248 | 0.000 | 0.174 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
| IDCARDNUM | ID | 0.994 | 0.865 | 0.858 | 0.900 | 0.994 | 0.953 | 0.919 | 0.710 | 0.881 | 0.936 | 0.904 | 0.717 | 0.836 | 0.258 | 0.939 | 0.932 | 0.990 | 0.990 | 0.711 | 0.211 | 0.864 | 0.660 | 0.630 | 0.132 | 0.067 | 0.906 | 0.774 | 0.847 | 0.821 | 0.926 | 0.843 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.974 | 0.620 | 0.000 | 0.704 | 0.734 | 0.681 | 0.074 | 0.444 | 0.053 | 0.092 | 0.019 | 0.340 | 0.000 | 0.061 | 0.025 | 0.012 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
| LICENSEPLATENUM | ID | 1.000 | 0.981 | 0.992 | 0.972 | 1.000 | 0.674 | 0.915 | 0.985 | 0.832 | 0.630 | 0.698 | 0.320 | 0.917 | 0.194 | 0.861 | 0.861 | 0.951 | 0.902 | 0.750 | 0.127 | 0.463 | 0.080 | 0.094 | 0.205 | 0.092 | 0.861 | 0.304 | 0.969 | 0.834 | 0.478 | 0.638 | 0.002 | 0.001 | 0.000 | 0.003 | 0.003 | 0.001 | 0.736 | 0.718 | 0.012 | 0.389 | 0.384 | 0.442 | 0.021 | 0.331 | 0.024 | 0.089 | 0.070 | 0.355 | 0.007 | 0.000 | 0.055 | 0.016 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
| NATIONALID | ID | 1.000 | 0.905 | 0.918 | 0.919 | 1.000 | 0.968 | 0.897 | 0.719 | 0.930 | 0.909 | 0.757 | 0.863 | 0.878 | 0.521 | 0.999 | 0.998 | 0.996 | 0.993 | 0.764 | 0.572 | 0.848 | 0.881 | 0.886 | 0.195 | 0.331 | 0.867 | 0.715 | 0.898 | 0.867 | 0.841 | 0.869 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.875 | 0.888 | 0.000 | 0.705 | 0.708 | 0.878 | 0.061 | 0.816 | 0.069 | 0.024 | 0.014 | 0.365 | 0.000 | 0.028 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
| ORGANIZATION | ORG | 0.560 | 0.987 | 0.987 | 0.987 | 0.000 | 1.000 | 1.000 | 1.000 | 0.293 | 0.293 | 0.187 | 0.240 | 1.000 | 0.187 | 0.000 | 0.000 | 0.307 | 0.000 | 0.747 | 0.360 | 0.347 | 0.173 | 0.213 | 0.853 | 0.853 | 0.187 | 0.440 | 0.853 | 0.667 | 0.347 | 0.000 | 0.320 | 0.320 | 1.000 | 1.000 | 1.000 | 0.320 | 0.000 | 0.827 | 0.773 | 0.533 | 0.307 | 0.000 | 1.000 | 0.387 | 0.000 | 1.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
| PASSPORTID | ID | 1.000 | 0.991 | 0.995 | 1.000 | 1.000 | 1.000 | 0.993 | 0.916 | 0.866 | 0.936 | 0.929 | 0.757 | 0.818 | 0.158 | 0.987 | 0.987 | 1.000 | 1.000 | 0.904 | 0.163 | 0.966 | 0.625 | 0.598 | 0.210 | 0.748 | 0.993 | 0.744 | 0.979 | 0.766 | 0.972 | 0.967 | 0.000 | 0.000 | 0.000 | 0.007 | 0.000 | 0.000 | 1.000 | 0.817 | 0.000 | 0.700 | 0.706 | 0.858 | 0.091 | 0.852 | 0.015 | 0.164 | 0.036 | 0.392 | 0.000 | 0.000 | 0.008 | 0.008 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
| PASSWORD | SECRET | 0.959 | 0.885 | 0.748 | 0.822 | 0.951 | 0.947 | 0.529 | 0.936 | 0.936 | 0.523 | 0.766 | 0.904 | 0.738 | 0.528 | 0.885 | 0.885 | 0.925 | 0.895 | 0.952 | 0.784 | 0.918 | 0.802 | 0.767 | 0.444 | 0.364 | 0.597 | 0.763 | 0.939 | 0.653 | 0.378 | 0.856 | 0.073 | 0.069 | 0.010 | 0.015 | 0.012 | 0.064 | 0.489 | 0.756 | 0.027 | 0.546 | 0.586 | 0.760 | 0.051 | 0.409 | 0.074 | 0.144 | 0.080 | 0.062 | 0.004 | 0.000 | 0.388 | 0.296 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
| PHONENUMBER | CONTACT | 1.000 | 0.999 | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | 0.984 | 0.992 | 0.750 | 0.993 | 1.000 | 0.757 | 0.999 | 0.999 | 1.000 | 1.000 | 0.997 | 0.908 | 0.997 | 0.968 | 0.958 | 0.851 | 1.000 | 1.000 | 0.976 | 0.997 | 0.939 | 0.886 | 0.990 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.976 | 0.994 | 0.000 | 0.926 | 0.937 | 0.994 | 0.043 | 0.971 | 0.197 | 0.000 | 0.007 | 0.169 | 0.000 | 0.810 | 0.005 | 0.000 | 0.007 | 0.008 | 0.000 | 0.000 | 0.000 |
| SECURITYTOKEN | SECRET | 0.989 | 0.893 | 0.920 | 0.876 | 0.989 | 0.967 | 0.946 | 0.990 | 0.937 | 0.871 | 0.834 | 0.659 | 0.949 | 0.783 | 0.811 | 0.811 | 0.980 | 0.956 | 0.942 | 0.714 | 0.918 | 0.026 | 0.026 | 0.665 | 0.842 | 0.871 | 0.872 | 0.914 | 0.989 | 0.789 | 0.980 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.628 | 0.898 | 0.000 | 0.597 | 0.583 | 0.934 | 0.130 | 0.777 | 0.272 | 0.176 | 0.077 | 0.177 | 0.000 | 0.374 | 0.186 | 0.172 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
| SSN | ID | 1.000 | 0.961 | 0.987 | 0.958 | 0.996 | 0.937 | 0.946 | 0.907 | 0.884 | 0.920 | 0.744 | 0.863 | 0.818 | 0.383 | 0.983 | 0.983 | 0.995 | 0.996 | 0.602 | 0.509 | 0.907 | 0.813 | 0.828 | 0.217 | 0.457 | 0.933 | 0.665 | 0.789 | 0.895 | 0.822 | 0.888 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.830 | 0.921 | 0.000 | 0.807 | 0.785 | 0.630 | 0.000 | 0.667 | 0.036 | 0.002 | 0.000 | 0.505 | 0.000 | 0.008 | 0.000 | 0.000 | 0.000 | 0.007 | 0.000 | 0.000 | 0.000 |
| STATE | ADDRESS | 0.998 | 0.987 | 0.964 | 0.969 | 0.937 | 0.963 | 0.986 | 0.958 | 0.916 | 0.713 | 0.871 | 0.608 | 0.891 | 0.852 | 0.295 | 0.297 | 0.287 | 0.231 | 0.929 | 0.602 | 0.285 | 0.601 | 0.605 | 0.842 | 0.699 | 0.856 | 0.384 | 0.915 | 0.753 | 0.437 | 0.207 | 0.737 | 0.718 | 0.987 | 0.991 | 0.989 | 0.687 | 0.347 | 0.547 | 0.937 | 0.578 | 0.578 | 0.138 | 0.999 | 0.598 | 0.483 | 0.744 | 0.418 | 0.252 | 0.274 | 0.000 | 0.061 | 0.029 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
| STREET | ADDRESS | 0.999 | 0.991 | 0.989 | 0.984 | 0.999 | 0.993 | 0.997 | 0.965 | 0.930 | 0.945 | 0.766 | 0.894 | 0.868 | 0.916 | 0.866 | 0.864 | 0.891 | 0.897 | 0.967 | 0.886 | 0.677 | 0.623 | 0.620 | 0.870 | 0.984 | 0.909 | 0.812 | 0.847 | 0.756 | 0.731 | 0.716 | 0.728 | 0.705 | 0.991 | 0.983 | 0.969 | 0.486 | 0.213 | 0.149 | 0.840 | 0.476 | 0.470 | 0.266 | 0.354 | 0.439 | 0.676 | 0.395 | 0.175 | 0.027 | 0.016 | 0.000 | 0.008 | 0.003 | 0.002 | 0.000 | 0.000 | 0.000 | 0.000 |
| SURNAME | PERSON | 0.988 | 0.978 | 0.977 | 0.967 | 0.972 | 0.907 | 0.981 | 0.916 | 0.957 | 0.897 | 0.804 | 0.893 | 0.741 | 0.830 | 0.885 | 0.884 | 0.773 | 0.782 | 0.206 | 0.863 | 0.878 | 0.912 | 0.907 | 0.825 | 0.330 | 0.151 | 0.807 | 0.105 | 0.614 | 0.566 | 0.770 | 0.921 | 0.922 | 0.972 | 0.967 | 0.959 | 0.905 | 0.777 | 0.376 | 0.931 | 0.384 | 0.359 | 0.875 | 0.723 | 0.069 | 0.509 | 0.899 | 0.319 | 0.103 | 0.214 | 0.006 | 0.099 | 0.112 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
| TAXNUM | ID | 1.000 | 0.862 | 0.872 | 0.857 | 1.000 | 0.984 | 0.945 | 0.868 | 0.843 | 0.923 | 0.559 | 0.682 | 0.867 | 0.343 | 0.962 | 0.962 | 0.994 | 0.996 | 0.826 | 0.255 | 0.840 | 0.598 | 0.574 | 0.292 | 0.034 | 0.939 | 0.525 | 0.898 | 0.820 | 0.894 | 0.875 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.912 | 0.714 | 0.000 | 0.643 | 0.679 | 0.683 | 0.071 | 0.636 | 0.046 | 0.073 | 0.094 | 0.303 | 0.000 | 0.008 | 0.016 | 0.008 | 0.000 | 0.010 | 0.000 | 0.000 | 0.000 |
| URL | NET | 0.982 | 0.801 | 0.834 | 0.812 | 0.968 | 0.357 | 0.510 | 0.304 | 0.881 | 0.804 | 0.377 | 0.518 | 0.351 | 0.467 | 0.394 | 0.387 | 0.702 | 0.780 | 0.989 | 0.874 | 0.315 | 0.835 | 0.816 | 0.060 | 0.084 | 0.004 | 0.070 | 0.917 | 0.960 | 0.383 | 0.027 | 0.027 | 0.028 | 0.089 | 0.113 | 0.047 | 0.029 | 0.153 | 0.868 | 0.033 | 0.387 | 0.380 | 0.024 | 0.064 | 0.502 | 0.018 | 0.014 | 0.032 | 0.093 | 0.000 | 0.999 | 0.002 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
| USERNAME | ACCOUNT | 1.000 | 1.000 | 0.968 | 1.000 | 1.000 | 1.000 | 0.901 | 0.979 | 0.959 | 0.684 | 0.826 | 0.927 | 0.856 | 0.688 | 0.793 | 0.804 | 0.751 | 0.699 | 0.979 | 0.763 | 0.921 | 0.260 | 0.340 | 0.450 | 0.095 | 0.935 | 0.636 | 0.959 | 0.727 | 0.307 | 0.615 | 0.271 | 0.237 | 0.284 | 0.121 | 0.015 | 0.163 | 0.692 | 0.964 | 0.181 | 0.785 | 0.764 | 0.895 | 0.019 | 0.873 | 0.292 | 0.079 | 0.079 | 0.307 | 0.013 | 0.107 | 0.095 | 0.064 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
| ZIP | ADDRESS | 1.000 | 1.000 | 0.996 | 0.998 | 1.000 | 0.997 | 0.832 | 0.999 | 0.877 | 0.978 | 0.939 | 0.995 | 0.469 | 0.789 | 0.874 | 0.869 | 0.937 | 0.955 | 0.801 | 0.843 | 0.909 | 0.781 | 0.768 | 0.610 | 0.826 | 0.319 | 0.791 | 0.759 | 0.552 | 0.870 | 0.823 | 0.081 | 0.070 | 0.053 | 0.015 | 0.015 | 0.059 | 0.428 | 0.580 | 0.031 | 0.471 | 0.486 | 0.296 | 0.586 | 0.683 | 0.493 | 0.069 | 0.022 | 0.212 | 0.037 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |

## char F1 by domain

| domain | n | pplx+cpu-int8 | nym-base | nym-small | nym-base+cpu-int8 | pplx | gliner2-fastino | bardsai-eu | gliner2-hivetrace-omni | gliner-nvidia | mmbert32k | gliner-stream-pii | openmed-multilingual | gliner2-large | kalyan-ettin | openai-base-onnx | openai-base | apararti | opf-kz-ru | nuner-zero | openmed-nemotron | opf-ru | gravitee-small | gravitee-small+cpu-int8 | gliner-multi-v21 | gliner25-fastino | gliner2-hivetrace-uni | traciora | gliner-urchade | pii-shield-onnx | mmbert32k+cpu-int8 | opf-ru-v2 | ner-ru-gherman | ner-ru-gherman-onnx | davlan-xlmr | davlan-mbert | davlan-mbert+cpu-int8 | ner-ru-gherman+cpu-int8 | ru-pii-ner | gliner-pii-edge | davlan-xlmr+cpu-int8 | ru-legal-ner+cpu-int8 | ru-legal-ner | gliner2-vladlinv | ner-ru-yqelz | gliner-pii-base | kalyan-ettin+cpu-int8 | stanza-ru | spacy-ru-lg | gliner-pii-edge+cpu-int8 | natasha | rules-ru | fef2-secret-ru | fef2-secret-ru+cpu-int8 | spacy-alrosait | gliner-pii-base+cpu-int8 | gliner-multi-v21+cpu-int8 | gliner-nvidia+cpu-int8 | gliner-urchade+cpu-int8 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Danish | 189 | 0.788 | 0.892 | 0.894 | 0.887 | 0.878 | 0.868 | 0.843 | 0.862 | 0.867 | 0.809 | 0.772 | 0.858 | 0.797 | 0.726 | 0.720 | 0.717 | 0.780 | 0.756 | 0.786 | 0.575 | 0.705 | 0.629 | 0.625 | 0.732 | 0.756 | 0.761 | 0.664 | 0.784 | 0.329 | 0.665 | 0.640 | 0.600 | 0.597 | 0.628 | 0.617 | 0.612 | 0.577 | 0.544 | 0.674 | 0.618 | 0.270 | 0.269 | 0.655 | 0.579 | 0.661 | 0.456 | 0.342 | 0.153 | 0.301 | 0.118 | 0.309 | 0.098 | 0.082 | 0.005 | 0.000 | 0.000 | 0.000 | 0.000 |
| Dutch | 199 | 0.759 | 0.899 | 0.905 | 0.901 | 0.880 | 0.864 | 0.855 | 0.872 | 0.872 | 0.781 | 0.773 | 0.758 | 0.777 | 0.720 | 0.796 | 0.795 | 0.807 | 0.790 | 0.798 | 0.677 | 0.756 | 0.566 | 0.560 | 0.712 | 0.736 | 0.772 | 0.728 | 0.793 | 0.302 | 0.611 | 0.708 | 0.586 | 0.584 | 0.653 | 0.639 | 0.635 | 0.549 | 0.587 | 0.677 | 0.626 | 0.232 | 0.232 | 0.646 | 0.572 | 0.659 | 0.453 | 0.339 | 0.159 | 0.331 | 0.128 | 0.180 | 0.121 | 0.081 | 0.004 | 0.002 | 0.000 | 0.000 | 0.000 |
| French | 200 | 0.744 | 0.917 | 0.918 | 0.919 | 0.869 | 0.884 | 0.855 | 0.880 | 0.858 | 0.838 | 0.860 | 0.877 | 0.835 | 0.796 | 0.802 | 0.800 | 0.816 | 0.811 | 0.778 | 0.739 | 0.784 | 0.738 | 0.740 | 0.676 | 0.784 | 0.766 | 0.769 | 0.764 | 0.487 | 0.685 | 0.782 | 0.528 | 0.518 | 0.620 | 0.614 | 0.607 | 0.473 | 0.741 | 0.655 | 0.541 | 0.266 | 0.267 | 0.650 | 0.546 | 0.702 | 0.630 | 0.309 | 0.257 | 0.279 | 0.139 | 0.273 | 0.134 | 0.077 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
| German | 204 | 0.753 | 0.902 | 0.904 | 0.900 | 0.858 | 0.891 | 0.835 | 0.892 | 0.859 | 0.795 | 0.794 | 0.869 | 0.806 | 0.774 | 0.799 | 0.799 | 0.816 | 0.804 | 0.732 | 0.688 | 0.740 | 0.742 | 0.734 | 0.685 | 0.762 | 0.763 | 0.713 | 0.778 | 0.380 | 0.672 | 0.738 | 0.609 | 0.606 | 0.636 | 0.630 | 0.619 | 0.567 | 0.683 | 0.616 | 0.606 | 0.226 | 0.228 | 0.666 | 0.544 | 0.635 | 0.540 | 0.250 | 0.189 | 0.287 | 0.114 | 0.274 | 0.175 | 0.138 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
| Spanish | 208 | 0.733 | 0.908 | 0.911 | 0.905 | 0.843 | 0.848 | 0.823 | 0.844 | 0.828 | 0.794 | 0.830 | 0.749 | 0.797 | 0.733 | 0.816 | 0.816 | 0.815 | 0.814 | 0.736 | 0.709 | 0.795 | 0.723 | 0.719 | 0.676 | 0.736 | 0.727 | 0.754 | 0.765 | 0.480 | 0.631 | 0.765 | 0.562 | 0.561 | 0.592 | 0.586 | 0.587 | 0.527 | 0.693 | 0.705 | 0.573 | 0.285 | 0.282 | 0.652 | 0.576 | 0.657 | 0.519 | 0.298 | 0.183 | 0.274 | 0.171 | 0.297 | 0.146 | 0.128 | 0.000 | 0.004 | 0.000 | 0.000 | 0.000 |

## Missed at thresholds 0.5 / 0.3 / 0.2 / 0.1

Only models that return a score. A lower threshold keeps more spans: fewer misses, more masking.

| model | 0.5 | 0.3 | 0.2 | 0.1 |
|---|---|---|---|---|
| nym-base | 188 (2.5%) | 177 (2.3%) | 177 (2.3%) | 177 (2.3%) |
| nym-small | 204 (2.7%) | 201 (2.7%) | 201 (2.7%) | 201 (2.7%) |
| nym-base+cpu-int8 | 258 (3.4%) | 243 (3.2%) | 243 (3.2%) | 243 (3.2%) |
| gliner2-fastino | 319 (4.2%) | 233 (3.1%) | 197 (2.6%) | 152 (2.0%) |
| bardsai-eu | 368 (4.9%) | 231 (3.1%) | 206 (2.7%) | 200 (2.6%) |
| gliner2-hivetrace-omni | 440 (5.8%) | 233 (3.1%) | 164 (2.2%) | 130 (1.7%) |
| gliner-nvidia | 512 (6.8%) | 361 (4.8%) | 303 (4.0%) | 243 (3.2%) |
| mmbert32k | 976 (12.9%) | 430 (5.7%) | 383 (5.1%) | 379 (5.0%) |
| gliner-stream-pii | 1251 (16.5%) | 747 (9.9%) | 535 (7.1%) | 311 (4.1%) |
| openmed-multilingual | 1263 (16.7%) | 1124 (14.9%) | 1124 (14.9%) | 1124 (14.9%) |
| gliner2-large | 1495 (19.8%) | 759 (10.0%) | 504 (6.7%) | 298 (3.9%) |
| kalyan-ettin | 1516 (20.0%) | 1090 (14.4%) | 1062 (14.0%) | 1062 (14.0%) |
| openai-base-onnx | 1663 (22.0%) | 1653 (21.9%) | 1653 (21.9%) | 1653 (21.9%) |
| openai-base | 1677 (22.2%) | 1665 (22.0%) | 1665 (22.0%) | 1665 (22.0%) |
| apararti | 1683 (22.2%) | 1653 (21.9%) | 1653 (21.9%) | 1653 (21.9%) |
| opf-kz-ru | 1698 (22.4%) | 1678 (22.2%) | 1677 (22.2%) | 1677 (22.2%) |
| nuner-zero | 1857 (24.5%) | 1189 (15.7%) | 892 (11.8%) | 560 (7.4%) |
| openmed-nemotron | 1889 (25.0%) | 1509 (19.9%) | 1481 (19.6%) | 1481 (19.6%) |
| opf-ru | 1919 (25.4%) | 1812 (24.0%) | 1808 (23.9%) | 1807 (23.9%) |
| gravitee-small | 1970 (26.0%) | 1805 (23.9%) | 1804 (23.8%) | 1804 (23.8%) |
| gravitee-small+cpu-int8 | 2020 (26.7%) | 1852 (24.5%) | 1852 (24.5%) | 1852 (24.5%) |
| gliner-multi-v21 | 2090 (27.6%) | 819 (10.8%) | 452 (6.0%) | 311 (4.1%) |
| gliner25-fastino | 2248 (29.7%) | 2040 (27.0%) | 1983 (26.2%) | 1887 (24.9%) |
| gliner2-hivetrace-uni | 2356 (31.1%) | 1530 (20.2%) | 1105 (14.6%) | 608 (8.0%) |
| traciora | 2371 (31.3%) | 2331 (30.8%) | 2327 (30.8%) | 2327 (30.8%) |
| gliner-urchade | 2400 (31.7%) | 2049 (27.1%) | 1884 (24.9%) | 1700 (22.5%) |
| pii-shield-onnx | 2409 (31.8%) | 2116 (28.0%) | 2111 (27.9%) | 2111 (27.9%) |
| mmbert32k+cpu-int8 | 2469 (32.6%) | 1055 (13.9%) | 705 (9.3%) | 639 (8.4%) |
| opf-ru-v2 | 2718 (35.9%) | 2683 (35.5%) | 2682 (35.5%) | 2682 (35.5%) |
| ner-ru-gherman | 2965 (39.2%) | 2680 (35.4%) | 2662 (35.2%) | 2658 (35.1%) |
| ner-ru-gherman-onnx | 2989 (39.5%) | 2703 (35.7%) | 2685 (35.5%) | 2677 (35.4%) |
| davlan-xlmr | 3106 (41.1%) | 3096 (40.9%) | 3096 (40.9%) | 3096 (40.9%) |
| davlan-mbert | 3249 (42.9%) | 3248 (42.9%) | 3248 (42.9%) | 3248 (42.9%) |
| davlan-mbert+cpu-int8 | 3295 (43.6%) | 3288 (43.5%) | 3288 (43.5%) | 3288 (43.5%) |
| ner-ru-gherman+cpu-int8 | 3300 (43.6%) | 2837 (37.5%) | 2780 (36.7%) | 2772 (36.6%) |
| gliner-pii-edge | 3426 (45.3%) | 681 (9.0%) | 147 (1.9%) | 16 (0.2%) |
| davlan-xlmr+cpu-int8 | 3464 (45.8%) | 3356 (44.4%) | 3350 (44.3%) | 3350 (44.3%) |
| ru-legal-ner+cpu-int8 | 3481 (46.0%) | 1647 (21.8%) | 1432 (18.9%) | 1404 (18.6%) |
| ru-legal-ner | 3482 (46.0%) | 1779 (23.5%) | 1569 (20.7%) | 1541 (20.4%) |
| gliner2-vladlinv | 3556 (47.0%) | 3169 (41.9%) | 3006 (39.7%) | 2732 (36.1%) |
| ner-ru-yqelz | 3616 (47.8%) | 3503 (46.3%) | 3503 (46.3%) | 3503 (46.3%) |
| gliner-pii-base | 3782 (50.0%) | 1320 (17.4%) | 509 (6.7%) | 194 (2.6%) |
| kalyan-ettin+cpu-int8 | 3904 (51.6%) | 3320 (43.9%) | 3279 (43.3%) | 3276 (43.3%) |
| gliner-pii-edge+cpu-int8 | 6145 (81.2%) | 792 (10.5%) | 200 (2.6%) | 36 (0.5%) |
| fef2-secret-ru | 7135 (94.3%) | 7129 (94.2%) | 7129 (94.2%) | 7129 (94.2%) |
| fef2-secret-ru+cpu-int8 | 7208 (95.3%) | 7199 (95.2%) | 7199 (95.2%) | 7199 (95.2%) |
| gliner-pii-base+cpu-int8 | 7562 (100.0%) | 7083 (93.6%) | 4644 (61.4%) | 753 (10.0%) |
| gliner-multi-v21+cpu-int8 | 7565 (100.0%) | 7565 (100.0%) | 7565 (100.0%) | 7562 (100.0%) |
| gliner-nvidia+cpu-int8 | 7565 (100.0%) | 7565 (100.0%) | 7565 (100.0%) | 7561 (99.9%) |
| gliner-urchade+cpu-int8 | 7565 (100.0%) | 7565 (100.0%) | 7548 (99.8%) | 7187 (95.0%) |

Notes: the unit of counting is one gold span after boundary normalization - the same unit in `missed`, in the group table, in the threshold table and in the ensembles; two annotations that share boundaries but not type are two spans, so an empty answer misses 100%. Gold and predicted boundaries are normalized the same way (whitespace, word boundaries, merge of adjacent same-label pieces). `dropped spans` - predicted intervals outside the text: rejected, not clipped to fit. A run made before `run.py` started clipping span ends to the piece it fed the model can carry them; `meta.clipped` counts what the clipping fixes in newer runs. Zero-shot models get the taxonomy of the set as labels (the `labels` field of `meta.json`). `train` - the model was trained on the source of this set (its slices and corrupted copies included): its row is not comparable with the others and stays out of the pooled numbers; an empty `contaminated` list in `benchmark/models.toml` means no evidence of overlap was found, not proof of none. `hidden` - gold spans every character of which is covered; a touched span counts as detected, not as hidden. missed % and char F1 carry 95% bootstrap intervals over rows (1000 resamples, the same rows for every model of the set). The group of a gold type comes from `BENCH/<set>/meta.json`; a model label with no group falls into OTHER. ms/row is a median and includes chunking of long texts; hardware and batch are in the meta line of the prediction file.