# kiji-en - en / pii (1000 rows, 7623 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 0.182 (81.8% 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 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| nym-small | **172** | 2.3% [1.9%, 2.6%] | 96.1% | 0.907 [0.901, 0.913] | 0.860 | 0.960 | 0.737 | 0.909 | 0.436 | 0/0 | 0 | 0 | 70 | - |  |
| nym-base | **209** | 2.7% [2.4%, 3.2%] | 95.4% | 0.901 [0.894, 0.907] | 0.859 | 0.947 | 0.750 | 0.903 | 0.437 | 0/0 | 0 | 0 | 4 | 308M |  |
| gliner2-fastino | **245** | 3.2% [2.8%, 3.7%] | 96.1% | 0.878 [0.870, 0.885] | 0.821 | 0.943 | 0.757 | 0.897 | 0.813 | 0/0 | 0 | 8 | 6 | 307M |  |
| gliner-nvidia | **277** | 3.6% [3.1%, 4.3%] | 93.1% | 0.874 [0.867, 0.882] | 0.833 | 0.919 | 0.721 | 0.890 | 0.804 | 0/0 | 0 | 0 | 12 | 445M |  |
| pplx | **310** | 4.1% [3.5%, 4.6%] | 95.7% | 0.863 [0.857, 0.870] | 0.801 | 0.935 | 0.409 | 0.915 | 0.811 | 0/0 | 0 | 0 | 54 | 596M |  |
| bardsai-eu | **319** | 4.2% [3.7%, 4.7%] | 89.7% | 0.842 [0.836, 0.849] | 0.776 | 0.921 | 0.387 | 0.891 | 0.431 | 0/0 | 0 | 0 | 774 | - |  |
| gravitee-small | **324** | 4.3% [3.8%, 4.8%] | 95.3% | 0.923 [0.917, 0.928] | 0.909 | 0.937 | 0.600 | 0.938 | 0.873 | 0/0 | 0 | 0 | 3 | 29M |  |
| gliner2-hivetrace-omni | **332** | 4.4% [3.9%, 4.8%] | 95.0% | 0.883 [0.876, 0.891] | 0.842 | 0.928 | 0.776 | 0.897 | 0.838 | 0/0 | 0 | 3 | 6 | 307M |  |
| openmed-multilingual | **368** | 4.8% [4.3%, 5.4%] | 93.3% | 0.894 [0.888, 0.899] | 0.849 | 0.944 | 0.729 | 0.878 | 0.864 | 0/0 | 0 | 0 | 32 | 1.4B |  |
| openmed-nemotron | **627** | 8.2% [7.6%, 8.9%] | 89.1% | 0.856 [0.850, 0.863] | 0.825 | 0.890 | 0.682 | 0.866 | 0.813 | 0/0 | 0 | 0 | 18 | 1.4B |  |
| kalyan-ettin | **681** | 8.9% [8.3%, 9.7%] | 86.0% | 0.845 [0.838, 0.852] | 0.828 | 0.864 | 0.646 | 0.872 | 0.806 | 0/0 | 0 | 0 | 6 | 68M |  |
| gliner-stream-pii | **758** | 9.9% [9.2%, 10.7%] | 85.8% | 0.849 [0.841, 0.857] | 0.854 | 0.844 | 0.795 | 0.870 | 0.772 | 0/0 | 0 | 0 | 29 | 677M |  |
| mmbert32k | **979** | 12.8% [12.1%, 13.7%] | 77.4% | 0.845 [0.839, 0.852] | 0.855 | 0.836 | 0.325 | 0.853 | 0.458 | 0/0 | 0 | 0 | 4 | 308M |  |
| gliner2-large | **1012** | 13.3% [12.4%, 14.1%] | 85.9% | 0.850 [0.842, 0.858] | 0.812 | 0.892 | 0.693 | 0.846 | 0.787 | 0/0 | 0 | 17 | 8 | 486M |  |
| apararti | **1114** | 14.6% [13.6%, 15.6%] | 84.8% | 0.825 [0.818, 0.833] | 0.810 | 0.841 | 0.386 | 0.886 | 0.766 | 0/0 | 0 | 0 | 64 | 1.4B |  |
| openai-base | **1158** | 15.2% [14.1%, 16.3%] | 84.4% | 0.810 [0.802, 0.819] | 0.798 | 0.823 | 0.375 | 0.876 | 0.760 | 0/0 | 0 | 0 | 71 | 1.4B |  |
| opf-kz-ru | **1239** | 16.3% [15.2%, 17.3%] | 82.8% | 0.824 [0.816, 0.832] | 0.821 | 0.827 | 0.368 | 0.878 | 0.758 | 0/0 | 0 | 0 | 62 | 1.4B |  |
| traciora | **1531** | 20.1% [19.0%, 21.2%] | 76.6% | 0.794 [0.785, 0.802] | 0.825 | 0.765 | 0.312 | 0.865 | 0.742 | 0/0 | 0 | 0 | 272 | 1.4B |  |
| opf-ru-v2 | **1777** | 23.3% [22.1%, 24.6%] | 73.6% | 0.788 [0.779, 0.797] | 0.845 | 0.738 | 0.303 | 0.849 | 0.742 | 0/0 | 0 | 0 | 270 | 1.4B |  |
| nuner-zero | **1817** | 23.8% [22.8%, 24.7%] | 73.7% | 0.804 [0.796, 0.813] | 0.782 | 0.828 | 0.583 | 0.787 | 0.741 | 0/0 | 0 | 0 | 7 | 449M |  |
| opf-ru | **1883** | 24.7% [23.7%, 25.7%] | 69.1% | 0.783 [0.775, 0.792] | 0.877 | 0.708 | 0.405 | 0.817 | 0.732 | 0/0 | 0 | 0 | 43 | 1.4B |  |
| gliner25-fastino | **1906** | 25.0% [23.9%, 26.1%] | 74.1% | 0.802 [0.793, 0.811] | 0.862 | 0.751 | 0.657 | 0.808 | 0.752 | 0/0 | 0 | 0 | 3 | 287M |  |
| stanza-ru | **2086** | 27.4% [26.3%, 28.4%] | 70.4% | 0.458 [0.448, 0.468] | 0.359 | 0.631 | 0.217 | 0.725 | 0.018 | 0/0 | 0 | 0 | 38 | - |  |
| gliner2-hivetrace-uni | **2101** | 27.6% [26.4%, 28.8%] | 71.4% | 0.788 [0.779, 0.797] | 0.851 | 0.733 | 0.596 | 0.787 | 0.603 | 0/0 | 0 | 0 | 13 | 147M |  |
| gliner-multi-v21 | **2310** | 30.3% [29.1%, 31.6%] | 67.3% | 0.668 [0.656, 0.679] | 0.735 | 0.612 | 0.405 | 0.744 | 0.594 | 0/0 | 0 | 0 | 4 | 289M |  |
| ru-pii-ner | **2470** | 32.4% [30.8%, 34.1%] | 66.8% | 0.715 [0.703, 0.726] | 0.773 | 0.665 | 0.377 | 0.753 | 0.330 | 0/0 | 0 | 0 | 90 | 358M |  |
| gliner-pii-base | **2772** | 36.4% [35.3%, 37.5%] | 62.1% | 0.750 [0.740, 0.759] | 0.836 | 0.679 | 0.568 | 0.724 | 0.683 | 0/0 | 0 | 0 | 3 | 166M |  |
| davlan-xlmr | **2863** | 37.6% [36.7%, 38.5%] | 60.0% | 0.646 [0.635, 0.656] | 0.779 | 0.551 | 0.431 | 0.705 | 0.245 | 0/0 | 0 | 0 | 4 | 277M |  |
| gliner-urchade | **2928** | 38.4% [37.2%, 39.6%] | 60.1% | 0.746 [0.736, 0.757] | 0.787 | 0.709 | 0.473 | 0.699 | 0.677 | 0/0 | 0 | 0 | 4 | 289M |  |
| ner-ru-gherman | **2930** | 38.4% [37.5%, 39.4%] | 56.3% | 0.591 [0.580, 0.601] | 0.798 | 0.470 | 0.619 | 0.698 | 0.251 | 0/0 | 0 | 0 | 4 | 177M |  |
| gliner-pii-edge | **3004** | 39.4% [38.1%, 40.6%] | 59.1% | 0.718 [0.708, 0.728] | 0.805 | 0.647 | 0.599 | 0.697 | 0.647 | 0/0 | 0 | 0 | 151 | 45M |  |
| davlan-mbert | **3205** | 42.0% [41.3%, 42.9%] | 56.3% | 0.630 [0.619, 0.640] | 0.777 | 0.530 | 0.452 | 0.670 | 0.245 | 0/0 | 0 | 0 | 4 | 177M |  |
| gliner2-vladlinv | **3414** | 44.8% [43.1%, 46.5%] | 54.9% | 0.663 [0.650, 0.676] | 0.843 | 0.547 | 0.398 | 0.673 | 0.493 | 0/0 | 0 | 0 | 6 | 287M |  |
| ner-ru-yqelz | **3426** | 44.9% [44.0%, 45.9%] | 51.6% | 0.575 [0.563, 0.585] | 0.733 | 0.473 | 0.339 | 0.662 | 0.187 | 0/0 | 0 | 0 | 5 | 559M |  |
| pii-shield-onnx | **4310** | 56.5% [55.2%, 57.9%] | 37.6% | 0.637 [0.623, 0.650] | 0.899 | 0.494 | 0.384 | 0.581 | 0.456 | 0/0 | 0 | 0 | 1619 | - |  |
| ru-legal-ner | **4821** | 63.2% [61.7%, 64.9%] | 28.0% | 0.554 [0.536, 0.570] | 0.882 | 0.404 | 0.230 | 0.525 | 0.321 | 0/0 | 0 | 0 | 3 | 29M |  |
| spacy-ru-lg | **6048** | 79.3% [78.1%, 80.5%] | 20.5% | 0.243 [0.231, 0.256] | 0.353 | 0.186 | 0.126 | 0.324 | 0.033 | 0/0 | 0 | 0 | 6 | - |  |
| fef2-secret-ru | **6466** | 84.8% [83.6%, 86.1%] | 13.7% | 0.310 [0.289, 0.331] | 0.920 | 0.186 | 0.154 | 0.262 | 0.122 | 0/0 | 0 | 0 | 4 | 177M |  |
| natasha | **6541** | 85.8% [84.9%, 86.7%] | 14.0% | 0.203 [0.190, 0.216] | 0.744 | 0.118 | 0.147 | 0.243 | 0.050 | 0/0 | 0 | 0 | 3 | - |  |
| rules-ru | **7003** | 91.9% [91.2%, 92.5%] | 7.8% | 0.317 [0.298, 0.338] | 0.969 | 0.190 | 0.125 | 0.150 | 0.315 | 0/0 | 0 | 0 | 0 | - |  |
| spacy-alrosait | **7623** | 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 | 5 | - |  |

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 ≈ gliner2-fastino; gliner2-fastino ≈ gliner-nvidia; gliner-nvidia ≈ pplx; pplx ≈ bardsai-eu; bardsai-eu ≈ gravitee-small; gravitee-small ≈ gliner2-hivetrace-omni; gliner2-hivetrace-omni ≈ openmed-multilingual; openmed-nemotron ≈ kalyan-ettin; kalyan-ettin ≈ gliner-stream-pii; mmbert32k ≈ gliner2-large; apararti ≈ openai-base; opf-ru-v2 ≈ nuner-zero; nuner-zero ≈ opf-ru; opf-ru ≈ gliner25-fastino; stanza-ru ≈ gliner2-hivetrace-uni; gliner-multi-v21 ≈ ru-pii-ner; gliner-pii-base ≈ davlan-xlmr; davlan-xlmr ≈ gliner-urchade; gliner-urchade ≈ ner-ru-gherman; ner-ru-gherman ≈ gliner-pii-edge; gliner2-vladlinv ≈ ner-ru-yqelz; fef2-secret-ru ≈ natasha

## Missed by group

| model | PERSON | ADDRESS | CONTACT | ID | NET | ACCOUNT | SECRET | ORG | DATE |
|---|---|---|---|---|---|---|---|---|---|
| spans in gold | 1926 | 3541 | 484 | 878 | 161 | 46 | 211 | 241 | 135 |
| nym-small | 39 (2.0%) | 53 (1.5%) | 3 (0.6%) | 23 (2.6%) | 5 (3.1%) | 0 (0.0%) | 28 (13.3%) | 18 (7.5%) | 3 (2.2%) |
| nym-base | 27 (1.4%) | 40 (1.1%) | 4 (0.8%) | 70 (8.0%) | 8 (5.0%) | 1 (2.2%) | 41 (19.4%) | 12 (5.0%) | 6 (4.4%) |
| gliner2-fastino | 83 (4.3%) | 37 (1.0%) | 0 (0.0%) | 11 (1.3%) | 98 (60.9%) | 0 (0.0%) | 2 (0.9%) | 0 (0.0%) | 14 (10.4%) |
| gliner-nvidia | 37 (1.9%) | 113 (3.2%) | 10 (2.1%) | 78 (8.9%) | 16 (9.9%) | 3 (6.5%) | 13 (6.2%) | 1 (0.4%) | 6 (4.4%) |
| pplx | 9 (0.5%) | 99 (2.8%) | 0 (0.0%) | 8 (0.9%) | 17 (10.6%) | 0 (0.0%) | 3 (1.4%) | 174 (72.2%) | 0 (0.0%) |
| bardsai-eu | 10 (0.5%) | 181 (5.1%) | 3 (0.6%) | 59 (6.7%) | 7 (4.3%) | 1 (2.2%) | 26 (12.3%) | 3 (1.2%) | 29 (21.5%) |
| gravitee-small | 20 (1.0%) | 41 (1.2%) | 0 (0.0%) | 95 (10.8%) | 0 (0.0%) | 43 (93.5%) | 114 (54.0%) | 7 (2.9%) | 4 (3.0%) |
| gliner2-hivetrace-omni | 42 (2.2%) | 107 (3.0%) | 7 (1.4%) | 46 (5.2%) | 103 (64.0%) | 0 (0.0%) | 2 (0.9%) | 10 (4.1%) | 15 (11.1%) |
| openmed-multilingual | 29 (1.5%) | 204 (5.8%) | 0 (0.0%) | 44 (5.0%) | 19 (11.8%) | 2 (4.3%) | 3 (1.4%) | 67 (27.8%) | 0 (0.0%) |
| openmed-nemotron | 22 (1.1%) | 191 (5.4%) | 1 (0.2%) | 293 (33.4%) | 21 (13.0%) | 1 (2.2%) | 33 (15.6%) | 65 (27.0%) | 0 (0.0%) |
| kalyan-ettin | 99 (5.1%) | 194 (5.5%) | 10 (2.1%) | 260 (29.6%) | 44 (27.3%) | 2 (4.3%) | 27 (12.8%) | 41 (17.0%) | 4 (3.0%) |
| gliner-stream-pii | 176 (9.1%) | 356 (10.1%) | 35 (7.2%) | 43 (4.9%) | 55 (34.2%) | 0 (0.0%) | 19 (9.0%) | 54 (22.4%) | 20 (14.8%) |
| mmbert32k | 73 (3.8%) | 747 (21.1%) | 1 (0.2%) | 65 (7.4%) | 3 (1.9%) | 12 (26.1%) | 45 (21.3%) | 33 (13.7%) | 0 (0.0%) |
| gliner2-large | 247 (12.8%) | 620 (17.5%) | 0 (0.0%) | 13 (1.5%) | 101 (62.7%) | 0 (0.0%) | 1 (0.5%) | 2 (0.8%) | 28 (20.7%) |
| apararti | 162 (8.4%) | 659 (18.6%) | 0 (0.0%) | 24 (2.7%) | 49 (30.4%) | 1 (2.2%) | 12 (5.7%) | 207 (85.9%) | 0 (0.0%) |
| openai-base | 70 (3.6%) | 727 (20.5%) | 1 (0.2%) | 36 (4.1%) | 68 (42.2%) | 0 (0.0%) | 36 (17.1%) | 220 (91.3%) | 0 (0.0%) |
| opf-kz-ru | 220 (11.4%) | 727 (20.5%) | 0 (0.0%) | 23 (2.6%) | 40 (24.8%) | 3 (6.5%) | 16 (7.6%) | 210 (87.1%) | 0 (0.0%) |
| traciora | 352 (18.3%) | 665 (18.8%) | 9 (1.9%) | 86 (9.8%) | 100 (62.1%) | 4 (8.7%) | 28 (13.3%) | 201 (83.4%) | 86 (63.7%) |
| opf-ru-v2 | 432 (22.4%) | 805 (22.7%) | 5 (1.0%) | 56 (6.4%) | 121 (75.2%) | 4 (8.7%) | 10 (4.7%) | 213 (88.4%) | 131 (97.0%) |
| nuner-zero | 1545 (80.2%) | 162 (4.6%) | 2 (0.4%) | 86 (9.8%) | 0 (0.0%) | 2 (4.3%) | 15 (7.1%) | 2 (0.8%) | 3 (2.2%) |
| opf-ru | 206 (10.7%) | 1278 (36.1%) | 5 (1.0%) | 82 (9.3%) | 81 (50.3%) | 1 (2.2%) | 6 (2.8%) | 221 (91.7%) | 3 (2.2%) |
| gliner25-fastino | 889 (46.2%) | 395 (11.2%) | 5 (1.0%) | 296 (33.7%) | 136 (84.5%) | 31 (67.4%) | 38 (18.0%) | 91 (37.8%) | 25 (18.5%) |
| stanza-ru | 25 (1.3%) | 649 (18.3%) | 416 (86.0%) | 572 (65.1%) | 157 (97.5%) | 37 (80.4%) | 114 (54.0%) | 4 (1.7%) | 112 (83.0%) |
| gliner2-hivetrace-uni | 1205 (62.6%) | 454 (12.8%) | 26 (5.4%) | 130 (14.8%) | 130 (80.7%) | 0 (0.0%) | 35 (16.6%) | 103 (42.7%) | 18 (13.3%) |
| gliner-multi-v21 | 298 (15.5%) | 764 (21.6%) | 326 (67.4%) | 616 (70.2%) | 145 (90.1%) | 13 (28.3%) | 73 (34.6%) | 50 (20.7%) | 25 (18.5%) |
| ru-pii-ner | 126 (6.5%) | 1907 (53.9%) | 29 (6.0%) | 35 (4.0%) | 94 (58.4%) | 2 (4.3%) | 40 (19.0%) | 237 (98.3%) | 0 (0.0%) |
| gliner-pii-base | 1671 (86.8%) | 722 (20.4%) | 23 (4.8%) | 118 (13.4%) | 101 (62.7%) | 3 (6.5%) | 23 (10.9%) | 42 (17.4%) | 69 (51.1%) |
| davlan-xlmr | 8 (0.4%) | 1078 (30.4%) | 418 (86.4%) | 877 (99.9%) | 99 (61.5%) | 37 (80.4%) | 210 (99.5%) | 1 (0.4%) | 135 (100.0%) |
| gliner-urchade | 1805 (93.7%) | 975 (27.5%) | 17 (3.5%) | 87 (9.9%) | 8 (5.0%) | 2 (4.3%) | 18 (8.5%) | 0 (0.0%) | 16 (11.9%) |
| ner-ru-gherman | 36 (1.9%) | 970 (27.4%) | 346 (71.5%) | 868 (98.9%) | 110 (68.3%) | 38 (82.6%) | 209 (99.1%) | 218 (90.5%) | 135 (100.0%) |
| gliner-pii-edge | 1008 (52.3%) | 1643 (46.4%) | 20 (4.1%) | 162 (18.5%) | 18 (11.2%) | 0 (0.0%) | 53 (25.1%) | 26 (10.8%) | 74 (54.8%) |
| davlan-mbert | 21 (1.1%) | 1353 (38.2%) | 464 (95.9%) | 873 (99.4%) | 107 (66.5%) | 41 (89.1%) | 208 (98.6%) | 3 (1.2%) | 135 (100.0%) |
| gliner2-vladlinv | 208 (10.8%) | 2626 (74.2%) | 44 (9.1%) | 125 (14.2%) | 143 (88.8%) | 1 (2.2%) | 21 (10.0%) | 240 (99.6%) | 6 (4.4%) |
| ner-ru-yqelz | 506 (26.3%) | 1116 (31.5%) | 460 (95.0%) | 770 (87.7%) | 150 (93.2%) | 44 (95.7%) | 191 (90.5%) | 55 (22.8%) | 134 (99.3%) |
| pii-shield-onnx | 1502 (78.0%) | 1981 (55.9%) | 28 (5.8%) | 516 (58.8%) | 11 (6.8%) | 29 (63.0%) | 23 (10.9%) | 212 (88.0%) | 8 (5.9%) |
| ru-legal-ner | 1249 (64.8%) | 2795 (78.9%) | 13 (2.7%) | 377 (42.9%) | 48 (29.8%) | 27 (58.7%) | 126 (59.7%) | 147 (61.0%) | 39 (28.9%) |
| spacy-ru-lg | 1230 (63.9%) | 2831 (79.9%) | 462 (95.5%) | 857 (97.6%) | 147 (91.3%) | 42 (91.3%) | 203 (96.2%) | 145 (60.2%) | 131 (97.0%) |
| fef2-secret-ru | 1574 (81.7%) | 3222 (91.0%) | 227 (46.9%) | 860 (97.9%) | 149 (92.5%) | 29 (63.0%) | 95 (45.0%) | 175 (72.6%) | 135 (100.0%) |
| natasha | 1514 (78.6%) | 2906 (82.1%) | 482 (99.6%) | 878 (100.0%) | 161 (100.0%) | 45 (97.8%) | 210 (99.5%) | 210 (87.1%) | 135 (100.0%) |
| rules-ru | 1924 (99.9%) | 3539 (99.9%) | 137 (28.3%) | 855 (97.4%) | 0 (0.0%) | 43 (93.5%) | 158 (74.9%) | 212 (88.0%) | 135 (100.0%) |
| spacy-alrosait | 1926 (100.0%) | 3541 (100.0%) | 484 (100.0%) | 878 (100.0%) | 161 (100.0%) | 46 (100.0%) | 211 (100.0%) | 241 (100.0%) | 135 (100.0%) |

## Char recall by gold type

| type | group | nym-small | nym-base | gliner2-fastino | gliner-nvidia | pplx | bardsai-eu | gravitee-small | gliner2-hivetrace-omni | openmed-multilingual | openmed-nemotron | kalyan-ettin | gliner-stream-pii | mmbert32k | gliner2-large | apararti | openai-base | opf-kz-ru | traciora | opf-ru-v2 | nuner-zero | opf-ru | gliner25-fastino | stanza-ru | gliner2-hivetrace-uni | gliner-multi-v21 | ru-pii-ner | gliner-pii-base | davlan-xlmr | gliner-urchade | ner-ru-gherman | gliner-pii-edge | davlan-mbert | gliner2-vladlinv | ner-ru-yqelz | pii-shield-onnx | ru-legal-ner | spacy-ru-lg | fef2-secret-ru | natasha | rules-ru | spacy-alrosait |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| BUILDINGNUM | ADDRESS | 0.995 | 0.990 | 0.965 | 0.977 | 0.994 | 0.874 | 0.952 | 0.961 | 0.936 | 0.888 | 0.901 | 0.863 | 0.978 | 0.516 | 0.953 | 0.925 | 0.947 | 0.921 | 0.923 | 0.937 | 0.907 | 0.819 | 0.541 | 0.887 | 0.661 | 0.484 | 0.749 | 0.231 | 0.713 | 0.527 | 0.295 | 0.016 | 0.387 | 0.115 | 0.325 | 0.151 | 0.054 | 0.010 | 0.000 | 0.005 | 0.000 |
| CITY | ADDRESS | 0.994 | 0.984 | 0.993 | 0.956 | 0.955 | 0.995 | 0.989 | 0.928 | 0.923 | 0.921 | 0.921 | 0.919 | 0.549 | 0.976 | 0.674 | 0.667 | 0.652 | 0.713 | 0.606 | 0.949 | 0.339 | 0.940 | 0.950 | 0.954 | 0.937 | 0.406 | 0.887 | 0.997 | 0.668 | 0.935 | 0.810 | 0.996 | 0.204 | 0.996 | 0.508 | 0.283 | 0.357 | 0.160 | 0.380 | 0.002 | 0.000 |
| COMPANYNAME | ORG | 0.908 | 0.944 | 1.000 | 0.991 | 0.260 | 0.988 | 0.964 | 0.963 | 0.640 | 0.671 | 0.730 | 0.779 | 0.749 | 0.993 | 0.113 | 0.081 | 0.118 | 0.151 | 0.111 | 0.982 | 0.066 | 0.630 | 0.985 | 0.570 | 0.791 | 0.015 | 0.814 | 0.991 | 0.996 | 0.042 | 0.877 | 0.981 | 0.002 | 0.720 | 0.085 | 0.329 | 0.399 | 0.267 | 0.133 | 0.122 | 0.000 |
| COUNTRY | ADDRESS | 0.984 | 0.993 | 1.000 | 0.929 | 0.811 | 0.981 | 0.933 | 1.000 | 0.959 | 0.984 | 0.911 | 0.916 | 0.399 | 1.000 | 0.488 | 0.472 | 0.410 | 0.510 | 0.474 | 0.928 | 0.185 | 0.942 | 0.996 | 0.966 | 0.934 | 0.382 | 0.913 | 0.994 | 0.934 | 0.991 | 0.791 | 0.991 | 0.061 | 1.000 | 0.557 | 0.097 | 0.536 | 0.288 | 0.409 | 0.000 | 0.000 |
| CREDITCARDNUMBER | ID | 0.955 | 0.888 | 1.000 | 0.966 | 1.000 | 0.753 | 1.000 | 0.974 | 0.913 | 0.832 | 0.902 | 0.396 | 0.895 | 0.980 | 0.934 | 0.904 | 0.932 | 0.847 | 0.903 | 0.966 | 0.842 | 0.825 | 0.053 | 0.141 | 0.314 | 1.000 | 0.923 | 0.000 | 0.966 | 0.000 | 0.958 | 0.000 | 0.906 | 0.347 | 0.318 | 0.663 | 0.000 | 0.000 | 0.000 | 0.255 | 0.000 |
| DATEOFBIRTH | DATE | 0.958 | 0.931 | 0.873 | 0.926 | 1.000 | 0.758 | 0.938 | 0.867 | 0.995 | 0.990 | 0.922 | 0.785 | 0.873 | 0.770 | 1.000 | 1.000 | 1.000 | 0.166 | 0.024 | 0.963 | 0.943 | 0.795 | 0.141 | 0.841 | 0.788 | 1.000 | 0.465 | 0.000 | 0.850 | 0.000 | 0.422 | 0.000 | 0.952 | 0.007 | 0.815 | 0.553 | 0.024 | 0.000 | 0.000 | 0.000 | 0.000 |
| DRIVERLICENSENUM | ID | 0.971 | 0.973 | 1.000 | 0.841 | 0.982 | 0.976 | 0.967 | 1.000 | 0.974 | 0.591 | 0.324 | 0.934 | 0.942 | 1.000 | 0.982 | 0.982 | 0.982 | 0.912 | 0.957 | 0.833 | 0.904 | 0.967 | 0.329 | 0.966 | 0.209 | 0.982 | 0.822 | 0.000 | 0.842 | 0.040 | 0.982 | 0.003 | 0.934 | 0.068 | 0.354 | 0.545 | 0.025 | 0.034 | 0.000 | 0.000 | 0.000 |
| EMAIL | CONTACT | 0.973 | 0.923 | 0.997 | 0.660 | 1.000 | 0.772 | 1.000 | 1.000 | 0.999 | 0.998 | 1.000 | 0.896 | 0.849 | 1.000 | 0.999 | 0.998 | 1.000 | 0.997 | 0.979 | 0.978 | 0.976 | 1.000 | 0.183 | 0.999 | 0.130 | 0.974 | 0.960 | 0.120 | 0.941 | 0.257 | 0.927 | 0.038 | 0.975 | 0.017 | 0.997 | 0.850 | 0.092 | 0.900 | 0.002 | 1.000 | 0.000 |
| FIRSTNAME | PERSON | 0.968 | 0.978 | 0.982 | 0.985 | 0.996 | 0.996 | 0.989 | 0.990 | 0.991 | 0.989 | 0.983 | 0.866 | 0.971 | 0.856 | 0.914 | 0.970 | 0.886 | 0.803 | 0.733 | 0.252 | 0.883 | 0.676 | 0.986 | 0.504 | 0.885 | 0.944 | 0.201 | 0.993 | 0.047 | 0.988 | 0.694 | 0.986 | 0.902 | 0.787 | 0.243 | 0.309 | 0.335 | 0.160 | 0.224 | 0.002 | 0.000 |
| IBAN | ID | 0.997 | 0.998 | 1.000 | 0.943 | 0.998 | 0.893 | 0.998 | 0.990 | 0.993 | 0.552 | 0.667 | 0.759 | 0.940 | 0.944 | 0.994 | 0.997 | 0.995 | 0.997 | 0.968 | 0.935 | 0.900 | 0.946 | 0.329 | 0.453 | 0.864 | 0.989 | 0.808 | 0.000 | 0.935 | 0.021 | 0.627 | 0.010 | 0.957 | 0.387 | 0.742 | 0.765 | 0.049 | 0.024 | 0.000 | 0.079 | 0.000 |
| IDCARDNUM | ID | 0.979 | 0.866 | 0.979 | 0.944 | 1.000 | 0.985 | 0.104 | 0.939 | 0.933 | 0.243 | 0.278 | 0.970 | 0.951 | 0.968 | 1.000 | 0.986 | 1.000 | 0.946 | 0.954 | 0.902 | 0.873 | 0.653 | 0.307 | 0.867 | 0.116 | 1.000 | 0.812 | 0.000 | 0.889 | 0.000 | 0.475 | 0.003 | 0.859 | 0.111 | 0.245 | 0.440 | 0.000 | 0.046 | 0.000 | 0.034 | 0.000 |
| LICENSEPLATENUM | ID | 0.960 | 0.967 | 0.971 | 0.916 | 0.979 | 0.861 | 0.972 | 0.979 | 0.979 | 0.941 | 0.858 | 0.916 | 0.540 | 0.979 | 0.979 | 0.979 | 0.973 | 0.846 | 0.826 | 0.916 | 0.735 | 0.379 | 0.231 | 0.817 | 0.401 | 0.873 | 0.886 | 0.000 | 0.921 | 0.003 | 0.739 | 0.006 | 0.763 | 0.043 | 0.237 | 0.194 | 0.015 | 0.008 | 0.000 | 0.000 | 0.000 |
| NATIONALID | ID | 0.937 | 0.910 | 0.950 | 0.871 | 0.960 | 0.947 | 0.980 | 0.667 | 0.960 | 0.957 | 0.802 | 0.885 | 0.923 | 0.980 | 0.960 | 0.948 | 0.960 | 0.802 | 0.884 | 0.891 | 0.855 | 0.599 | 0.227 | 0.980 | 0.145 | 0.940 | 0.849 | 0.020 | 0.871 | 0.000 | 0.928 | 0.000 | 0.906 | 0.105 | 0.296 | 0.482 | 0.052 | 0.035 | 0.000 | 0.023 | 0.000 |
| ORGANIZATION | ORG | 1.000 | 1.000 | 1.000 | 1.000 | 0.000 | 1.000 | 0.647 | 1.000 | 0.490 | 0.667 | 0.000 | 0.490 | 0.137 | 1.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 1.000 | 0.137 | 0.490 | 0.804 | 1.000 | 1.000 | 0.000 | 1.000 | 1.000 | 1.000 | 0.137 | 1.000 | 1.000 | 0.000 | 1.000 | 0.000 | 0.000 | 0.490 | 0.000 | 0.000 | 0.000 | 0.000 |
| PASSPORTID | ID | 0.989 | 0.955 | 1.000 | 0.956 | 0.989 | 0.989 | 0.997 | 0.990 | 0.954 | 0.239 | 0.574 | 0.980 | 0.947 | 1.000 | 0.989 | 0.989 | 0.989 | 0.981 | 0.964 | 0.964 | 0.954 | 0.935 | 0.596 | 1.000 | 0.096 | 0.989 | 0.964 | 0.000 | 0.964 | 0.009 | 0.905 | 0.000 | 0.954 | 0.053 | 0.150 | 0.455 | 0.027 | 0.018 | 0.000 | 0.000 | 0.000 |
| PASSWORD | SECRET | 0.916 | 0.946 | 0.981 | 0.972 | 0.982 | 0.664 | 0.991 | 0.976 | 0.991 | 0.972 | 0.964 | 0.832 | 0.510 | 0.968 | 0.969 | 0.958 | 0.920 | 0.865 | 0.866 | 0.935 | 0.959 | 0.654 | 0.355 | 0.705 | 0.453 | 0.732 | 0.864 | 0.007 | 0.907 | 0.017 | 0.864 | 0.023 | 0.843 | 0.106 | 0.733 | 0.384 | 0.022 | 0.911 | 0.001 | 0.030 | 0.000 |
| PHONENUMBER | CONTACT | 1.000 | 1.000 | 1.000 | 0.978 | 1.000 | 0.981 | 1.000 | 0.974 | 1.000 | 0.986 | 0.881 | 0.868 | 0.976 | 1.000 | 1.000 | 0.995 | 1.000 | 0.918 | 0.993 | 0.981 | 0.996 | 0.969 | 0.040 | 0.894 | 0.347 | 0.882 | 0.933 | 0.000 | 0.948 | 0.000 | 0.985 | 0.000 | 0.844 | 0.032 | 0.828 | 0.935 | 0.000 | 0.071 | 0.000 | 0.465 | 0.000 |
| REGION | ADDRESS | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | 0.000 | 1.000 | 1.000 | 0.000 | 1.000 | 1.000 | 0.000 | 1.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 1.000 | 0.000 | 1.000 | 0.000 | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | 0.000 | 1.000 | 0.000 | 1.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
| SECURITYTOKEN | SECRET | 0.794 | 0.678 | 0.995 | 0.915 | 0.994 | 0.969 | 0.000 | 1.000 | 0.963 | 0.693 | 0.739 | 0.940 | 0.872 | 1.000 | 0.921 | 0.709 | 0.921 | 0.849 | 0.987 | 0.921 | 0.965 | 0.973 | 0.368 | 0.934 | 0.780 | 0.812 | 0.908 | 0.000 | 0.905 | 0.000 | 0.630 | 0.000 | 0.958 | 0.075 | 0.990 | 0.324 | 0.044 | 0.227 | 0.000 | 0.460 | 0.000 |
| SSN | ID | 1.000 | 0.979 | 1.000 | 0.895 | 1.000 | 0.949 | 0.990 | 0.971 | 0.981 | 0.902 | 0.960 | 0.928 | 0.942 | 0.990 | 0.989 | 0.989 | 0.993 | 0.842 | 0.895 | 0.881 | 0.921 | 0.605 | 0.085 | 0.993 | 0.283 | 0.970 | 0.885 | 0.000 | 0.885 | 0.000 | 0.971 | 0.000 | 0.765 | 0.008 | 0.406 | 0.726 | 0.003 | 0.010 | 0.000 | 0.000 | 0.000 |
| STATE | ADDRESS | 0.962 | 0.984 | 0.972 | 0.909 | 0.949 | 0.980 | 0.984 | 0.966 | 0.872 | 0.925 | 0.831 | 0.946 | 0.503 | 0.935 | 0.603 | 0.621 | 0.521 | 0.642 | 0.536 | 0.873 | 0.214 | 0.765 | 0.966 | 0.882 | 0.730 | 0.549 | 0.566 | 0.983 | 0.831 | 0.837 | 0.620 | 0.959 | 0.150 | 0.996 | 0.613 | 0.220 | 0.528 | 0.193 | 0.524 | 0.000 | 0.000 |
| STREET | ADDRESS | 0.993 | 0.992 | 0.993 | 0.961 | 1.000 | 0.994 | 0.990 | 0.987 | 0.997 | 0.980 | 0.995 | 0.746 | 0.963 | 0.951 | 0.978 | 0.943 | 0.967 | 0.936 | 0.927 | 0.975 | 0.831 | 0.912 | 0.998 | 0.973 | 0.776 | 0.424 | 0.744 | 0.987 | 0.669 | 0.802 | 0.205 | 0.987 | 0.268 | 0.657 | 0.327 | 0.146 | 0.075 | 0.008 | 0.015 | 0.000 | 0.000 |
| SURNAME | PERSON | 0.994 | 0.997 | 0.914 | 0.973 | 0.992 | 0.992 | 0.988 | 0.958 | 0.978 | 0.992 | 0.903 | 0.954 | 0.959 | 0.881 | 0.931 | 0.963 | 0.889 | 0.839 | 0.851 | 0.107 | 0.909 | 0.349 | 0.986 | 0.183 | 0.788 | 0.931 | 0.043 | 0.998 | 0.081 | 0.973 | 0.201 | 0.991 | 0.873 | 0.669 | 0.181 | 0.424 | 0.394 | 0.212 | 0.211 | 0.001 | 0.000 |
| TAXNUM | ID | 0.952 | 0.745 | 0.965 | 0.886 | 0.990 | 0.897 | 0.978 | 0.946 | 0.926 | 0.752 | 0.868 | 0.836 | 0.897 | 0.953 | 0.990 | 0.934 | 0.979 | 0.724 | 0.895 | 0.848 | 0.780 | 0.119 | 0.126 | 0.968 | 0.332 | 0.907 | 0.835 | 0.000 | 0.856 | 0.000 | 0.623 | 0.000 | 0.677 | 0.108 | 0.337 | 0.532 | 0.013 | 0.013 | 0.000 | 0.000 | 0.000 |
| URL | NET | 0.776 | 0.737 | 0.327 | 0.803 | 0.888 | 0.609 | 0.998 | 0.291 | 0.870 | 0.862 | 0.565 | 0.561 | 0.825 | 0.306 | 0.671 | 0.541 | 0.710 | 0.299 | 0.116 | 0.986 | 0.255 | 0.108 | 0.018 | 0.085 | 0.040 | 0.361 | 0.285 | 0.156 | 0.882 | 0.122 | 0.805 | 0.145 | 0.056 | 0.032 | 0.847 | 0.458 | 0.096 | 0.035 | 0.000 | 0.945 | 0.000 |
| USERNAME | ACCOUNT | 1.000 | 0.965 | 1.000 | 0.919 | 1.000 | 0.963 | 0.062 | 1.000 | 0.948 | 0.977 | 0.952 | 1.000 | 0.739 | 1.000 | 0.981 | 1.000 | 0.942 | 0.923 | 0.934 | 0.930 | 0.979 | 0.319 | 0.182 | 1.000 | 0.656 | 0.971 | 0.896 | 0.186 | 0.930 | 0.201 | 1.000 | 0.118 | 0.967 | 0.041 | 0.404 | 0.414 | 0.097 | 0.385 | 0.017 | 0.068 | 0.000 |
| ZIP | ADDRESS | 0.998 | 0.999 | 0.999 | 0.951 | 1.000 | 0.878 | 0.999 | 0.994 | 0.990 | 0.952 | 0.933 | 0.877 | 0.970 | 0.759 | 0.946 | 0.912 | 0.950 | 0.905 | 0.925 | 0.946 | 0.871 | 0.925 | 0.603 | 0.558 | 0.583 | 0.581 | 0.822 | 0.129 | 0.697 | 0.050 | 0.811 | 0.012 | 0.342 | 0.498 | 0.415 | 0.231 | 0.012 | 0.028 | 0.036 | 0.000 | 0.000 |

## char F1 by domain

| domain | n | nym-small | nym-base | gliner2-fastino | gliner-nvidia | pplx | bardsai-eu | gravitee-small | gliner2-hivetrace-omni | openmed-multilingual | openmed-nemotron | kalyan-ettin | gliner-stream-pii | mmbert32k | gliner2-large | apararti | openai-base | opf-kz-ru | traciora | opf-ru-v2 | nuner-zero | opf-ru | gliner25-fastino | stanza-ru | gliner2-hivetrace-uni | gliner-multi-v21 | ru-pii-ner | gliner-pii-base | davlan-xlmr | gliner-urchade | ner-ru-gherman | gliner-pii-edge | davlan-mbert | gliner2-vladlinv | ner-ru-yqelz | pii-shield-onnx | ru-legal-ner | spacy-ru-lg | fef2-secret-ru | natasha | rules-ru | spacy-alrosait |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Australia | 137 | 0.882 | 0.884 | 0.858 | 0.839 | 0.837 | 0.831 | 0.897 | 0.862 | 0.875 | 0.836 | 0.820 | 0.850 | 0.815 | 0.837 | 0.807 | 0.799 | 0.805 | 0.778 | 0.778 | 0.763 | 0.793 | 0.792 | 0.407 | 0.796 | 0.671 | 0.730 | 0.727 | 0.603 | 0.723 | 0.578 | 0.705 | 0.599 | 0.682 | 0.569 | 0.632 | 0.444 | 0.196 | 0.215 | 0.206 | 0.253 | 0.000 |
| Canada | 135 | 0.901 | 0.894 | 0.882 | 0.877 | 0.860 | 0.853 | 0.920 | 0.886 | 0.893 | 0.851 | 0.827 | 0.826 | 0.824 | 0.854 | 0.822 | 0.796 | 0.812 | 0.783 | 0.778 | 0.788 | 0.746 | 0.793 | 0.473 | 0.805 | 0.716 | 0.692 | 0.756 | 0.670 | 0.761 | 0.624 | 0.698 | 0.658 | 0.631 | 0.619 | 0.630 | 0.406 | 0.233 | 0.187 | 0.252 | 0.238 | 0.000 |
| Ireland | 137 | 0.895 | 0.898 | 0.861 | 0.860 | 0.872 | 0.848 | 0.914 | 0.872 | 0.905 | 0.872 | 0.853 | 0.863 | 0.833 | 0.846 | 0.822 | 0.800 | 0.808 | 0.786 | 0.802 | 0.768 | 0.783 | 0.783 | 0.457 | 0.789 | 0.685 | 0.716 | 0.719 | 0.679 | 0.711 | 0.630 | 0.694 | 0.676 | 0.639 | 0.579 | 0.604 | 0.501 | 0.212 | 0.238 | 0.225 | 0.243 | 0.000 |
| New Zealand | 136 | 0.868 | 0.857 | 0.835 | 0.811 | 0.833 | 0.813 | 0.880 | 0.840 | 0.864 | 0.809 | 0.816 | 0.835 | 0.814 | 0.806 | 0.797 | 0.786 | 0.794 | 0.765 | 0.769 | 0.752 | 0.758 | 0.786 | 0.426 | 0.773 | 0.658 | 0.705 | 0.704 | 0.623 | 0.674 | 0.593 | 0.662 | 0.620 | 0.657 | 0.546 | 0.630 | 0.467 | 0.239 | 0.203 | 0.229 | 0.254 | 0.000 |
| United Kingdom | 153 | 0.906 | 0.894 | 0.883 | 0.881 | 0.853 | 0.840 | 0.915 | 0.891 | 0.893 | 0.858 | 0.859 | 0.851 | 0.839 | 0.849 | 0.817 | 0.809 | 0.823 | 0.788 | 0.783 | 0.812 | 0.784 | 0.800 | 0.436 | 0.805 | 0.719 | 0.711 | 0.754 | 0.663 | 0.760 | 0.603 | 0.721 | 0.631 | 0.661 | 0.610 | 0.621 | 0.461 | 0.222 | 0.199 | 0.211 | 0.273 | 0.000 |
| United States | 302 | 0.943 | 0.934 | 0.910 | 0.918 | 0.890 | 0.853 | 0.960 | 0.912 | 0.911 | 0.880 | 0.867 | 0.859 | 0.887 | 0.877 | 0.852 | 0.838 | 0.858 | 0.823 | 0.801 | 0.867 | 0.806 | 0.827 | 0.507 | 0.774 | 0.609 | 0.724 | 0.789 | 0.638 | 0.790 | 0.556 | 0.766 | 0.610 | 0.683 | 0.549 | 0.668 | 0.724 | 0.295 | 0.490 | 0.154 | 0.442 | 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-small | 172 (2.3%) | 157 (2.1%) | 157 (2.1%) | 157 (2.1%) |
| nym-base | 209 (2.7%) | 197 (2.6%) | 197 (2.6%) | 197 (2.6%) |
| gliner2-fastino | 245 (3.2%) | 177 (2.3%) | 152 (2.0%) | 130 (1.7%) |
| gliner-nvidia | 277 (3.6%) | 257 (3.4%) | 247 (3.2%) | 237 (3.1%) |
| bardsai-eu | 319 (4.2%) | 210 (2.8%) | 201 (2.6%) | 200 (2.6%) |
| gravitee-small | 324 (4.3%) | 313 (4.1%) | 313 (4.1%) | 313 (4.1%) |
| gliner2-hivetrace-omni | 332 (4.4%) | 194 (2.5%) | 158 (2.1%) | 130 (1.7%) |
| openmed-multilingual | 368 (4.8%) | 249 (3.3%) | 245 (3.2%) | 245 (3.2%) |
| openmed-nemotron | 627 (8.2%) | 513 (6.7%) | 510 (6.7%) | 510 (6.7%) |
| kalyan-ettin | 681 (8.9%) | 367 (4.8%) | 342 (4.5%) | 340 (4.5%) |
| gliner-stream-pii | 758 (9.9%) | 427 (5.6%) | 309 (4.1%) | 180 (2.4%) |
| mmbert32k | 979 (12.8%) | 457 (6.0%) | 390 (5.1%) | 389 (5.1%) |
| gliner2-large | 1012 (13.3%) | 621 (8.1%) | 499 (6.5%) | 374 (4.9%) |
| apararti | 1114 (14.6%) | 1099 (14.4%) | 1099 (14.4%) | 1099 (14.4%) |
| openai-base | 1158 (15.2%) | 1154 (15.1%) | 1154 (15.1%) | 1154 (15.1%) |
| opf-kz-ru | 1239 (16.3%) | 1224 (16.1%) | 1224 (16.1%) | 1224 (16.1%) |
| traciora | 1531 (20.1%) | 1507 (19.8%) | 1505 (19.7%) | 1505 (19.7%) |
| opf-ru-v2 | 1777 (23.3%) | 1763 (23.1%) | 1762 (23.1%) | 1762 (23.1%) |
| nuner-zero | 1817 (23.8%) | 1331 (17.5%) | 1016 (13.3%) | 718 (9.4%) |
| opf-ru | 1883 (24.7%) | 1808 (23.7%) | 1805 (23.7%) | 1805 (23.7%) |
| gliner25-fastino | 1906 (25.0%) | 1680 (22.0%) | 1620 (21.3%) | 1532 (20.1%) |
| gliner2-hivetrace-uni | 2101 (27.6%) | 1228 (16.1%) | 827 (10.8%) | 481 (6.3%) |
| gliner-multi-v21 | 2310 (30.3%) | 832 (10.9%) | 444 (5.8%) | 276 (3.6%) |
| gliner-pii-base | 2772 (36.4%) | 1454 (19.1%) | 731 (9.6%) | 280 (3.7%) |
| davlan-xlmr | 2863 (37.6%) | 2842 (37.3%) | 2841 (37.3%) | 2841 (37.3%) |
| gliner-urchade | 2928 (38.4%) | 2397 (31.4%) | 2228 (29.2%) | 1961 (25.7%) |
| ner-ru-gherman | 2930 (38.4%) | 2643 (34.7%) | 2631 (34.5%) | 2630 (34.5%) |
| gliner-pii-edge | 3004 (39.4%) | 565 (7.4%) | 70 (0.9%) | 1 (0.0%) |
| davlan-mbert | 3205 (42.0%) | 3202 (42.0%) | 3202 (42.0%) | 3202 (42.0%) |
| gliner2-vladlinv | 3414 (44.8%) | 3113 (40.8%) | 2960 (38.8%) | 2748 (36.0%) |
| ner-ru-yqelz | 3426 (44.9%) | 3344 (43.9%) | 3344 (43.9%) | 3344 (43.9%) |
| pii-shield-onnx | 4310 (56.5%) | 3277 (43.0%) | 3140 (41.2%) | 3128 (41.0%) |
| ru-legal-ner | 4821 (63.2%) | 3140 (41.2%) | 2618 (34.3%) | 2439 (32.0%) |
| fef2-secret-ru | 6466 (84.8%) | 6440 (84.5%) | 6440 (84.5%) | 6440 (84.5%) |

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.