# dialogpii-en - en / pii (147 rows, 3087 spans, 1 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 0.053 (94.7% 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 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| gliner2-fastino | **313** | 10.1% [8.2%, 12.1%] | 81.9% | 0.602 [0.580, 0.623] | 0.462 | 0.866 | 0.474 | 0.615 | 0.457 | 1/1 | 285 | 0 | 56 | 307M |  |
| gliner-nvidia+sent300 | **410** | 13.3% [11.2%, 15.3%] | 77.5% | 0.587 [0.568, 0.606] | 0.458 | 0.817 | 0.506 | 0.607 | 0.491 | 1/1 | 298 | 0 | 110 | 445M |  |
| gliner2-large | **437** | 14.2% [11.7%, 16.6%] | 78.0% | 0.571 [0.549, 0.592] | 0.441 | 0.809 | 0.460 | 0.588 | 0.473 | 1/1 | 264 | 0 | 63 | 486M |  |
| nuner-zero | **452** | 14.6% [13.0%, 16.3%] | 80.2% | 0.535 [0.516, 0.554] | 0.393 | 0.837 | 0.434 | 0.540 | 0.467 | 1/1 | 327 | 0 | 84 | 449M |  |
| gliner2-hivetrace-omni | **518** | 16.8% [14.8%, 18.6%] | 75.3% | 0.691 [0.670, 0.712] | 0.616 | 0.787 | 0.579 | 0.731 | 0.538 | 1/1 | 179 | 0 | 41 | 307M |  |
| stanza-ru | **525** | 17.0% [15.0%, 18.9%] | 78.8% | 0.187 [0.177, 0.196] | 0.107 | 0.744 | 0.085 | 0.469 | 0.009 | 1/1 | 1088 | 0 | 413 | - |  |
| bardsai-eu | **527** | 17.1% [15.0%, 19.0%] | 71.7% | 0.799 [0.782, 0.817] | 0.824 | 0.775 | 0.630 | 0.844 | 0.148 | 1/1 | 156 | 0 | 6470 | - |  |
| gliner-nvidia+ov100 | **531** | 17.2% [14.9%, 19.3%] | 73.3% | 0.635 [0.615, 0.655] | 0.537 | 0.777 | 0.559 | 0.669 | 0.537 | 1/1 | 230 | 0 | 88 | 445M |  |
| gliner-nvidia | **538** | 17.4% [15.1%, 19.6%] | 73.3% | 0.608 [0.588, 0.626] | 0.500 | 0.774 | 0.536 | 0.643 | 0.514 | 1/1 | 223 | 0 | 79 | 445M |  |
| gliner-multi-v21 | **544** | 17.6% [15.5%, 19.7%] | 75.1% | 0.531 [0.511, 0.548] | 0.404 | 0.773 | 0.387 | 0.498 | 0.402 | 1/1 | 332 | 0 | 34 | 289M |  |
| gliner-urchade | **570** | 18.5% [15.7%, 21.0%] | 75.4% | 0.507 [0.489, 0.525] | 0.371 | 0.800 | 0.379 | 0.491 | 0.401 | 1/1 | 367 | 0 | 44 | 289M |  |
| gliner-pii-base | **617** | 20.0% [17.4%, 22.3%] | 71.6% | 0.643 [0.623, 0.662] | 0.568 | 0.740 | 0.532 | 0.671 | 0.533 | 1/1 | 137 | 0 | 32 | 166M |  |
| gliner-pii-edge | **624** | 20.2% [17.5%, 23.0%] | 71.2% | 0.723 [0.700, 0.745] | 0.710 | 0.738 | 0.533 | 0.692 | 0.567 | 1/1 | 67 | 0 | 442 | 45M |  |
| openmed-multilingual | **660** | 21.4% [19.4%, 23.3%] | 56.2% | 0.603 [0.585, 0.620] | 0.565 | 0.645 | 0.429 | 0.659 | 0.483 | 1/1 | 171 | 0 | 200 | 1.4B |  |
| nym-small | **680** | 22.0% [19.5%, 24.5%] | 58.3% | 0.757 [0.733, 0.780] | 0.931 | 0.637 | 0.583 | 0.839 | 0.189 | 1/1 | 102 | 0 | 871 | - |  |
| nym-base+sent300 | **695** | 22.5% [20.0%, 25.0%] | 58.3% | 0.754 [0.730, 0.779] | 0.919 | 0.640 | 0.585 | 0.832 | 0.201 | 1/1 | 90 | 0 | 35 | 308M |  |
| gliner25-fastino+ov100 | **702** | 22.7% [20.2%, 25.0%] | 70.1% | 0.674 [0.651, 0.697] | 0.626 | 0.730 | 0.537 | 0.697 | 0.562 | 1/1 | 177 | 0 | 39 | 287M |  |
| davlan-xlmr | **718** | 23.3% [20.2%, 26.3%] | 66.4% | 0.780 [0.749, 0.811] | 0.972 | 0.651 | 0.725 | 0.857 | 0.518 | 1/1 | 62 | 0 | 36 | 277M |  |
| nym-base | **719** | 23.3% [20.7%, 25.8%] | 57.6% | 0.750 [0.725, 0.773] | 0.927 | 0.629 | 0.580 | 0.828 | 0.191 | 1/1 | 90 | 0 | 31 | 308M |  |
| gliner25-fastino+sent300 | **732** | 23.7% [21.4%, 25.9%] | 68.9% | 0.611 [0.588, 0.633] | 0.532 | 0.717 | 0.480 | 0.628 | 0.499 | 1/1 | 249 | 0 | 39 | 287M |  |
| gliner25-fastino | **742** | 24.0% [21.7%, 26.2%] | 69.0% | 0.655 [0.632, 0.679] | 0.604 | 0.716 | 0.526 | 0.685 | 0.546 | 1/1 | 216 | 0 | 24 | 287M |  |
| pplx | **744** | 24.1% [21.4%, 26.4%] | 71.0% | 0.761 [0.739, 0.784] | 0.822 | 0.708 | 0.632 | 0.822 | 0.682 | 1/1 | 150 | 0 | 481 | 596M |  |
| pplx+sent300 | **746** | 24.2% [21.8%, 26.4%] | 70.0% | 0.762 [0.741, 0.784] | 0.841 | 0.696 | 0.649 | 0.823 | 0.685 | 1/1 | 143 | 0 | 751 | 596M |  |
| pplx+ov100 | **752** | 24.4% [21.7%, 27.0%] | 71.4% | 0.756 [0.734, 0.778] | 0.810 | 0.709 | 0.628 | 0.817 | 0.678 | 1/1 | 163 | 0 | 602 | 596M |  |
| nym-base+ov100 | **754** | 24.4% [21.8%, 27.0%] | 56.6% | 0.738 [0.715, 0.762] | 0.924 | 0.615 | 0.574 | 0.820 | 0.180 | 1/1 | 88 | 0 | 40 | 308M |  |
| gravitee-small | **796** | 25.8% [23.0%, 28.3%] | 66.0% | 0.707 [0.687, 0.730] | 0.759 | 0.662 | 0.572 | 0.769 | 0.614 | 1/1 | 108 | 0 | 32 | 29M |  |
| gliner2-hivetrace-uni | **817** | 26.5% [24.2%, 28.5%] | 64.2% | 0.731 [0.714, 0.749] | 0.806 | 0.668 | 0.593 | 0.783 | 0.532 | 1/1 | 134 | 0 | 110 | 147M |  |
| davlan-mbert | **836** | 27.1% [23.9%, 30.2%] | 63.0% | 0.759 [0.727, 0.791] | 0.962 | 0.626 | 0.702 | 0.826 | 0.509 | 1/1 | 62 | 0 | 38 | 177M |  |
| openmed-nemotron | **906** | 29.3% [27.1%, 31.6%] | 51.1% | 0.523 [0.505, 0.542] | 0.491 | 0.561 | 0.406 | 0.614 | 0.436 | 1/1 | 233 | 0 | 226 | 1.4B |  |
| kalyan-ettin | **966** | 31.3% [28.8%, 33.6%] | 50.3% | 0.616 [0.595, 0.636] | 0.689 | 0.557 | 0.468 | 0.705 | 0.548 | 1/1 | 145 | 0 | 50 | 68M |  |
| gliner25-fastino+nochunk | **996** | 32.3% [28.8%, 35.5%] | 60.9% | 0.758 [0.732, 0.783] | 0.876 | 0.668 | 0.621 | 0.774 | 0.642 | 1/1 | 85 | 0 | 75 | 287M |  |
| ner-ru-yqelz | **1026** | 33.2% [30.5%, 36.0%] | 60.1% | 0.665 [0.639, 0.690] | 0.685 | 0.646 | 0.536 | 0.696 | 0.464 | 1/1 | 108 | 0 | 43 | 559M |  |
| ner-ru-gherman | **1053** | 34.1% [30.9%, 37.3%] | 46.1% | 0.634 [0.601, 0.665] | 0.983 | 0.468 | 0.530 | 0.787 | 0.104 | 1/1 | 55 | 0 | 38 | 177M |  |
| mmbert32k | **1095** | 35.5% [32.9%, 37.7%] | 46.6% | 0.543 [0.526, 0.563] | 0.558 | 0.529 | 0.380 | 0.595 | 0.380 | 1/1 | 129 | 0 | 41 | 308M |  |
| gliner2-vladlinv | **1179** | 38.2% [35.4%, 40.9%] | 55.9% | 0.633 [0.610, 0.658] | 0.878 | 0.495 | 0.553 | 0.748 | 0.468 | 1/1 | 101 | 0 | 47 | 287M |  |
| gliner-stream-pii | **1233** | 39.9% [36.9%, 42.9%] | 53.4% | 0.685 [0.664, 0.707] | 0.857 | 0.570 | 0.566 | 0.696 | 0.547 | 1/1 | 53 | 0 | 256 | 677M |  |
| ru-pii-ner | **1298** | 42.0% [39.2%, 44.7%] | 53.7% | 0.476 [0.454, 0.496] | 0.438 | 0.520 | 0.422 | 0.534 | 0.295 | 1/1 | 215 | 0 | 843 | 358M |  |
| openai-base | **1309** | 42.4% [39.3%, 45.5%] | 54.1% | 0.415 [0.394, 0.437] | 0.344 | 0.524 | 0.325 | 0.463 | 0.386 | 1/1 | 343 | 0 | 489 | 1.4B |  |
| apararti | **1334** | 43.2% [40.2%, 46.4%] | 52.0% | 0.455 [0.429, 0.479] | 0.395 | 0.536 | 0.339 | 0.492 | 0.415 | 1/1 | 243 | 0 | 460 | 1.4B |  |
| opf-ru | **1349** | 43.7% [41.0%, 46.6%] | 45.2% | 0.392 [0.370, 0.415] | 0.346 | 0.453 | 0.305 | 0.471 | 0.307 | 1/1 | 273 | 0 | 368 | 1.4B |  |
| opf-kz-ru | **1594** | 51.6% [48.3%, 55.0%] | 43.3% | 0.443 [0.414, 0.470] | 0.419 | 0.470 | 0.311 | 0.466 | 0.410 | 1/1 | 170 | 0 | 472 | 1.4B |  |
| traciora | **1655** | 53.6% [50.7%, 56.7%] | 40.3% | 0.580 [0.551, 0.607] | 0.854 | 0.439 | 0.393 | 0.613 | 0.522 | 1/1 | 118 | 0 | 373 | 1.4B |  |
| mmbert32k+nochunk | **1805** | 58.5% [55.3%, 61.5%] | 26.5% | 0.487 [0.460, 0.517] | 0.827 | 0.346 | 0.295 | 0.535 | 0.297 | 1/1 | 94 | 0 | 54 | 308M |  |
| opf-ru-v2+ov100 | **1904** | 61.7% [58.5%, 65.1%] | 33.6% | 0.480 [0.447, 0.512] | 0.695 | 0.366 | 0.294 | 0.510 | 0.450 | 1/1 | 144 | 0 | 4663 | 1.4B |  |
| opf-ru-v2+sent300 | **1925** | 62.4% [59.3%, 65.8%] | 32.8% | 0.483 [0.451, 0.514] | 0.706 | 0.367 | 0.293 | 0.504 | 0.446 | 1/1 | 116 | 0 | 60 | 1.4B |  |
| opf-ru-v2 | **1936** | 62.7% [59.2%, 66.1%] | 32.3% | 0.483 [0.451, 0.517] | 0.737 | 0.360 | 0.289 | 0.509 | 0.452 | 1/1 | 134 | 0 | 2209 | 1.4B |  |
| pii-shield-onnx | **2242** | 72.6% [70.5%, 74.9%] | 17.7% | 0.321 [0.294, 0.347] | 0.363 | 0.288 | 0.175 | 0.306 | 0.191 | 1/1 | 199 | 0 | 10722 | - |  |
| ru-legal-ner+sent300 | **2244** | 72.7% [70.6%, 74.6%] | 16.7% | 0.344 [0.319, 0.370] | 0.458 | 0.276 | 0.139 | 0.338 | 0.140 | 1/1 | 62 | 0 | 36 | 29M |  |
| ru-legal-ner+ov100 | **2583** | 83.7% [81.8%, 85.6%] | 7.5% | 0.267 [0.235, 0.297] | 0.633 | 0.169 | 0.092 | 0.258 | 0.146 | 1/1 | 20 | 0 | 35 | 29M |  |
| ru-legal-ner | **2632** | 85.3% [83.5%, 87.1%] | 6.5% | 0.244 [0.216, 0.269] | 0.596 | 0.153 | 0.082 | 0.235 | 0.129 | 1/1 | 39 | 0 | 32 | 29M |  |
| spacy-ru-lg | **2644** | 85.6% [83.7%, 87.4%] | 13.7% | 0.180 [0.157, 0.204] | 0.342 | 0.122 | 0.116 | 0.230 | 0.017 | 1/1 | 42 | 0 | 66 | - |  |
| natasha | **2778** | 90.0% [88.4%, 91.5%] | 9.6% | 0.124 [0.105, 0.145] | 0.554 | 0.070 | 0.140 | 0.174 | 0.047 | 1/1 | 16 | 0 | 19 | - |  |
| rules-ru | **2977** | 96.4% [95.5%, 97.5%] | 0.3% | 0.096 [0.068, 0.122] | 0.996 | 0.050 | 0.005 | 0.069 | 0.008 | 0/1 | 0 | 0 | 5 | - |  |
| fef2-secret-ru | **3032** | 98.2% [97.4%, 98.9%] | 1.2% | 0.038 [0.023, 0.056] | 0.908 | 0.019 | 0.019 | 0.035 | 0.026 | 0/1 | 0 | 0 | 44 | 177M |  |
| spacy-alrosait | **3083** | 99.9% [99.6%, 100.0%] | 0.1% | 0.002 [0.000, 0.005] | 0.081 | 0.001 | 0.000 | 0.003 | 0.000 | 0/1 | 0 | 0 | 64 | - |  |

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`: gliner-nvidia+sent300 ≈ gliner2-large; gliner2-large ≈ nuner-zero; gliner2-hivetrace-omni ≈ stanza-ru; stanza-ru ≈ bardsai-eu; bardsai-eu ≈ gliner-nvidia+ov100; gliner-nvidia+ov100 ≈ gliner-nvidia; gliner-nvidia ≈ gliner-multi-v21; gliner-multi-v21 ≈ gliner-urchade; gliner-urchade ≈ gliner-pii-base; gliner-pii-base ≈ gliner-pii-edge; gliner-pii-edge ≈ openmed-multilingual; openmed-multilingual ≈ nym-small; nym-small ≈ nym-base+sent300; nym-base+sent300 ≈ gliner25-fastino+ov100; gliner25-fastino+ov100 ≈ davlan-xlmr; davlan-xlmr ≈ nym-base; nym-base ≈ gliner25-fastino+sent300; gliner25-fastino+sent300 ≈ gliner25-fastino; gliner25-fastino ≈ pplx; pplx ≈ pplx+sent300; pplx+sent300 ≈ pplx+ov100; pplx+ov100 ≈ nym-base+ov100; nym-base+ov100 ≈ gravitee-small; gravitee-small ≈ gliner2-hivetrace-uni; gliner2-hivetrace-uni ≈ davlan-mbert; davlan-mbert ≈ openmed-nemotron; openmed-nemotron ≈ kalyan-ettin; kalyan-ettin ≈ gliner25-fastino+nochunk; gliner25-fastino+nochunk ≈ ner-ru-yqelz; ner-ru-yqelz ≈ ner-ru-gherman; ner-ru-gherman ≈ mmbert32k; gliner2-vladlinv ≈ gliner-stream-pii; gliner-stream-pii ≈ ru-pii-ner; ru-pii-ner ≈ openai-base; openai-base ≈ apararti; apararti ≈ opf-ru; mmbert32k+nochunk ≈ opf-ru-v2+ov100; opf-ru-v2+ov100 ≈ opf-ru-v2+sent300; opf-ru-v2+sent300 ≈ opf-ru-v2; pii-shield-onnx ≈ ru-legal-ner+sent300; ru-legal-ner ≈ spacy-ru-lg

## Missed by group

| model | PERSON | ADDRESS | CONTACT | ID | NET | ORG |
|---|---|---|---|---|---|---|
| spans in gold | 1558 | 483 | 200 | 260 | 7 | 579 |
| gliner2-fastino | 97 (6.2%) | 32 (6.6%) | 29 (14.5%) | 77 (29.6%) | 2 (28.6%) | 76 (13.1%) |
| gliner-nvidia+sent300 | 78 (5.0%) | 24 (5.0%) | 56 (28.0%) | 113 (43.5%) | 0 (0.0%) | 139 (24.0%) |
| gliner2-large | 119 (7.6%) | 56 (11.6%) | 42 (21.0%) | 133 (51.2%) | 2 (28.6%) | 85 (14.7%) |
| nuner-zero | 76 (4.9%) | 155 (32.1%) | 40 (20.0%) | 71 (27.3%) | 0 (0.0%) | 110 (19.0%) |
| gliner2-hivetrace-omni | 227 (14.6%) | 43 (8.9%) | 48 (24.0%) | 63 (24.2%) | 2 (28.6%) | 135 (23.3%) |
| stanza-ru | 90 (5.8%) | 54 (11.2%) | 156 (78.0%) | 116 (44.6%) | 7 (100.0%) | 102 (17.6%) |
| bardsai-eu | 146 (9.4%) | 65 (13.5%) | 53 (26.5%) | 94 (36.2%) | 2 (28.6%) | 167 (28.8%) |
| gliner-nvidia+ov100 | 133 (8.5%) | 33 (6.8%) | 55 (27.5%) | 143 (55.0%) | 0 (0.0%) | 167 (28.8%) |
| gliner-nvidia | 134 (8.6%) | 36 (7.5%) | 58 (29.0%) | 139 (53.5%) | 1 (14.3%) | 170 (29.4%) |
| gliner-multi-v21 | 141 (9.1%) | 77 (15.9%) | 64 (32.0%) | 150 (57.7%) | 6 (85.7%) | 106 (18.3%) |
| gliner-urchade | 141 (9.1%) | 75 (15.5%) | 40 (20.0%) | 195 (75.0%) | 0 (0.0%) | 119 (20.6%) |
| gliner-pii-base | 232 (14.9%) | 78 (16.1%) | 64 (32.0%) | 136 (52.3%) | 0 (0.0%) | 107 (18.5%) |
| gliner-pii-edge | 249 (16.0%) | 104 (21.5%) | 58 (29.0%) | 96 (36.9%) | 1 (14.3%) | 116 (20.0%) |
| openmed-multilingual | 115 (7.4%) | 101 (20.9%) | 51 (25.5%) | 88 (33.8%) | 1 (14.3%) | 304 (52.5%) |
| nym-small | 118 (7.6%) | 73 (15.1%) | 92 (46.0%) | 73 (28.1%) | 1 (14.3%) | 323 (55.8%) |
| nym-base+sent300 | 133 (8.5%) | 76 (15.7%) | 65 (32.5%) | 98 (37.7%) | 2 (28.6%) | 321 (55.4%) |
| gliner25-fastino+ov100 | 169 (10.8%) | 170 (35.2%) | 45 (22.5%) | 98 (37.7%) | 2 (28.6%) | 218 (37.7%) |
| davlan-xlmr | 92 (5.9%) | 101 (20.9%) | 163 (81.5%) | 249 (95.8%) | 6 (85.7%) | 107 (18.5%) |
| nym-base | 121 (7.8%) | 78 (16.1%) | 72 (36.0%) | 103 (39.6%) | 2 (28.6%) | 343 (59.2%) |
| gliner25-fastino+sent300 | 134 (8.6%) | 195 (40.4%) | 41 (20.5%) | 88 (33.8%) | 2 (28.6%) | 272 (47.0%) |
| gliner25-fastino | 158 (10.1%) | 185 (38.3%) | 45 (22.5%) | 100 (38.5%) | 2 (28.6%) | 252 (43.5%) |
| pplx | 52 (3.3%) | 125 (25.9%) | 45 (22.5%) | 46 (17.7%) | 0 (0.0%) | 476 (82.2%) |
| pplx+sent300 | 64 (4.1%) | 123 (25.5%) | 46 (23.0%) | 44 (16.9%) | 1 (14.3%) | 468 (80.8%) |
| pplx+ov100 | 52 (3.3%) | 133 (27.5%) | 44 (22.0%) | 42 (16.2%) | 1 (14.3%) | 480 (82.9%) |
| nym-base+ov100 | 133 (8.5%) | 84 (17.4%) | 70 (35.0%) | 99 (38.1%) | 2 (28.6%) | 366 (63.2%) |
| gravitee-small | 145 (9.3%) | 85 (17.6%) | 82 (41.0%) | 191 (73.5%) | 3 (42.9%) | 290 (50.1%) |
| gliner2-hivetrace-uni | 340 (21.8%) | 94 (19.5%) | 49 (24.5%) | 103 (39.6%) | 2 (28.6%) | 229 (39.6%) |
| davlan-mbert | 131 (8.4%) | 119 (24.6%) | 173 (86.5%) | 240 (92.3%) | 7 (100.0%) | 166 (28.7%) |
| openmed-nemotron | 144 (9.2%) | 118 (24.4%) | 68 (34.0%) | 158 (60.8%) | 3 (42.9%) | 415 (71.7%) |
| kalyan-ettin | 236 (15.1%) | 121 (25.1%) | 78 (39.0%) | 142 (54.6%) | 6 (85.7%) | 383 (66.1%) |
| gliner25-fastino+nochunk | 489 (31.4%) | 109 (22.6%) | 48 (24.0%) | 117 (45.0%) | 2 (28.6%) | 231 (39.9%) |
| ner-ru-yqelz | 566 (36.3%) | 68 (14.1%) | 132 (66.0%) | 205 (78.8%) | 1 (14.3%) | 54 (9.3%) |
| ner-ru-gherman | 126 (8.1%) | 140 (29.0%) | 158 (79.0%) | 249 (95.8%) | 6 (85.7%) | 374 (64.6%) |
| mmbert32k | 341 (21.9%) | 228 (47.2%) | 56 (28.0%) | 79 (30.4%) | 1 (14.3%) | 390 (67.4%) |
| gliner2-vladlinv | 104 (6.7%) | 298 (61.7%) | 76 (38.0%) | 135 (51.9%) | 7 (100.0%) | 559 (96.5%) |
| gliner-stream-pii | 652 (41.8%) | 96 (19.9%) | 95 (47.5%) | 120 (46.2%) | 3 (42.9%) | 267 (46.1%) |
| ru-pii-ner | 299 (19.2%) | 307 (63.6%) | 66 (33.0%) | 83 (31.9%) | 6 (85.7%) | 537 (92.7%) |
| openai-base | 307 (19.7%) | 330 (68.3%) | 63 (31.5%) | 69 (26.5%) | 7 (100.0%) | 533 (92.1%) |
| apararti | 387 (24.8%) | 316 (65.4%) | 58 (29.0%) | 57 (21.9%) | 3 (42.9%) | 513 (88.6%) |
| opf-ru | 329 (21.1%) | 340 (70.4%) | 65 (32.5%) | 90 (34.6%) | 4 (57.1%) | 521 (90.0%) |
| opf-kz-ru | 625 (40.1%) | 322 (66.7%) | 61 (30.5%) | 55 (21.2%) | 3 (42.9%) | 528 (91.2%) |
| traciora | 605 (38.8%) | 320 (66.3%) | 81 (40.5%) | 131 (50.4%) | 7 (100.0%) | 511 (88.3%) |
| mmbert32k+nochunk | 918 (58.9%) | 262 (54.2%) | 83 (41.5%) | 74 (28.5%) | 1 (14.3%) | 467 (80.7%) |
| opf-ru-v2+ov100 | 786 (50.4%) | 362 (74.9%) | 96 (48.0%) | 104 (40.0%) | 7 (100.0%) | 549 (94.8%) |
| opf-ru-v2+sent300 | 847 (54.4%) | 350 (72.5%) | 84 (42.0%) | 94 (36.2%) | 7 (100.0%) | 543 (93.8%) |
| opf-ru-v2 | 820 (52.6%) | 361 (74.7%) | 90 (45.0%) | 108 (41.5%) | 7 (100.0%) | 550 (95.0%) |
| pii-shield-onnx | 1251 (80.3%) | 313 (64.8%) | 55 (27.5%) | 174 (66.9%) | 3 (42.9%) | 446 (77.0%) |
| ru-legal-ner+sent300 | 1177 (75.5%) | 372 (77.0%) | 69 (34.5%) | 151 (58.1%) | 0 (0.0%) | 475 (82.0%) |
| ru-legal-ner+ov100 | 1366 (87.7%) | 429 (88.8%) | 79 (39.5%) | 152 (58.5%) | 5 (71.4%) | 552 (95.3%) |
| ru-legal-ner | 1407 (90.3%) | 426 (88.2%) | 83 (41.5%) | 170 (65.4%) | 4 (57.1%) | 542 (93.6%) |
| spacy-ru-lg | 1207 (77.5%) | 427 (88.4%) | 192 (96.0%) | 254 (97.7%) | 7 (100.0%) | 557 (96.2%) |
| natasha | 1375 (88.3%) | 418 (86.5%) | 199 (99.5%) | 256 (98.5%) | 7 (100.0%) | 523 (90.3%) |
| rules-ru | 1558 (100.0%) | 483 (100.0%) | 99 (49.5%) | 257 (98.8%) | 1 (14.3%) | 579 (100.0%) |
| fef2-secret-ru | 1530 (98.2%) | 476 (98.6%) | 190 (95.0%) | 256 (98.5%) | 7 (100.0%) | 573 (99.0%) |
| spacy-alrosait | 1554 (99.7%) | 483 (100.0%) | 200 (100.0%) | 260 (100.0%) | 7 (100.0%) | 579 (100.0%) |

## Char recall by gold type

| type | group | gliner2-fastino | gliner-nvidia+sent300 | gliner2-large | nuner-zero | gliner2-hivetrace-omni | stanza-ru | bardsai-eu | gliner-nvidia+ov100 | gliner-nvidia | gliner-multi-v21 | gliner-urchade | gliner-pii-base | gliner-pii-edge | openmed-multilingual | nym-small | nym-base+sent300 | gliner25-fastino+ov100 | davlan-xlmr | nym-base | gliner25-fastino+sent300 | gliner25-fastino | pplx | pplx+sent300 | pplx+ov100 | nym-base+ov100 | gravitee-small | gliner2-hivetrace-uni | davlan-mbert | openmed-nemotron | kalyan-ettin | gliner25-fastino+nochunk | ner-ru-yqelz | ner-ru-gherman | mmbert32k | gliner2-vladlinv | gliner-stream-pii | ru-pii-ner | openai-base | apararti | opf-ru | opf-kz-ru | traciora | mmbert32k+nochunk | opf-ru-v2+ov100 | opf-ru-v2+sent300 | opf-ru-v2 | pii-shield-onnx | ru-legal-ner+sent300 | ru-legal-ner+ov100 | ru-legal-ner | spacy-ru-lg | natasha | rules-ru | fef2-secret-ru | spacy-alrosait |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| CODE | ID | 0.664 | 0.550 | 0.454 | 0.679 | 0.659 | 0.422 | 0.569 | 0.428 | 0.433 | 0.363 | 0.256 | 0.430 | 0.533 | 0.522 | 0.573 | 0.503 | 0.559 | 0.017 | 0.473 | 0.583 | 0.552 | 0.784 | 0.795 | 0.817 | 0.493 | 0.199 | 0.533 | 0.028 | 0.297 | 0.321 | 0.505 | 0.179 | 0.023 | 0.575 | 0.406 | 0.386 | 0.626 | 0.676 | 0.709 | 0.478 | 0.729 | 0.427 | 0.571 | 0.512 | 0.546 | 0.503 | 0.327 | 0.403 | 0.382 | 0.340 | 0.026 | 0.006 | 0.010 | 0.018 | 0.000 |
| CODE_PHONE | CONTACT | 0.883 | 0.907 | 0.920 | 0.931 | 0.876 | 0.071 | 0.768 | 0.907 | 0.881 | 0.409 | 0.881 | 0.859 | 0.925 | 0.878 | 0.931 | 0.938 | 0.958 | 0.000 | 0.927 | 0.949 | 0.958 | 0.956 | 0.936 | 0.956 | 0.949 | 0.823 | 0.962 | 0.000 | 0.772 | 0.464 | 0.949 | 0.011 | 0.000 | 0.870 | 0.819 | 0.682 | 0.901 | 0.870 | 0.911 | 0.763 | 0.911 | 0.456 | 0.854 | 0.748 | 0.797 | 0.743 | 0.651 | 0.845 | 0.746 | 0.754 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
| CODE_URL | NET | 0.301 | 0.645 | 0.227 | 0.753 | 0.301 | 0.000 | 0.278 | 0.793 | 0.580 | 0.014 | 0.503 | 0.398 | 0.386 | 0.415 | 0.386 | 0.349 | 0.202 | 0.037 | 0.349 | 0.278 | 0.216 | 0.960 | 0.688 | 0.688 | 0.349 | 0.278 | 0.298 | 0.000 | 0.469 | 0.074 | 0.202 | 0.446 | 0.031 | 0.284 | 0.000 | 0.366 | 0.045 | 0.000 | 0.381 | 0.128 | 0.358 | 0.000 | 0.108 | 0.000 | 0.000 | 0.000 | 0.372 | 0.540 | 0.074 | 0.102 | 0.000 | 0.000 | 0.341 | 0.000 | 0.000 |
| LOC_CITY | ADDRESS | 0.986 | 0.971 | 0.925 | 0.549 | 0.919 | 0.901 | 0.928 | 0.961 | 0.961 | 0.921 | 0.903 | 0.841 | 0.877 | 0.778 | 0.891 | 0.893 | 0.667 | 0.914 | 0.908 | 0.574 | 0.613 | 0.779 | 0.771 | 0.766 | 0.891 | 0.844 | 0.889 | 0.921 | 0.748 | 0.741 | 0.825 | 0.948 | 0.803 | 0.485 | 0.381 | 0.845 | 0.341 | 0.175 | 0.203 | 0.184 | 0.191 | 0.216 | 0.381 | 0.146 | 0.166 | 0.145 | 0.382 | 0.222 | 0.089 | 0.106 | 0.130 | 0.198 | 0.000 | 0.010 | 0.000 |
| LOC_COUNTRY | ADDRESS | 0.657 | 0.879 | 0.682 | 0.542 | 0.631 | 0.815 | 0.486 | 0.818 | 0.804 | 0.659 | 0.621 | 0.710 | 0.561 | 0.505 | 0.533 | 0.537 | 0.444 | 0.343 | 0.519 | 0.343 | 0.437 | 0.292 | 0.350 | 0.276 | 0.533 | 0.586 | 0.484 | 0.325 | 0.783 | 0.614 | 0.561 | 0.776 | 0.393 | 0.086 | 0.035 | 0.488 | 0.150 | 0.019 | 0.000 | 0.049 | 0.000 | 0.035 | 0.042 | 0.019 | 0.019 | 0.035 | 0.332 | 0.100 | 0.047 | 0.030 | 0.042 | 0.098 | 0.000 | 0.030 | 0.000 |
| LOC_HOUSENUMBER | ADDRESS | 0.877 | 0.938 | 0.543 | 0.741 | 0.889 | 0.741 | 0.877 | 0.914 | 0.938 | 0.679 | 0.679 | 0.716 | 0.383 | 0.877 | 0.938 | 0.938 | 0.568 | 0.333 | 0.963 | 0.667 | 0.543 | 0.963 | 0.963 | 0.963 | 0.963 | 0.926 | 0.889 | 0.062 | 0.840 | 0.926 | 0.605 | 0.185 | 0.580 | 0.963 | 0.617 | 0.815 | 0.704 | 0.827 | 0.926 | 0.827 | 0.901 | 0.926 | 0.963 | 0.716 | 0.728 | 0.716 | 0.333 | 0.111 | 0.086 | 0.111 | 0.148 | 0.000 | 0.000 | 0.000 | 0.000 |
| LOC_OTHER | ADDRESS | 1.000 | 0.892 | 0.973 | 0.629 | 0.913 | 0.871 | 0.894 | 0.875 | 0.865 | 0.877 | 0.931 | 0.796 | 0.862 | 0.633 | 0.619 | 0.598 | 0.333 | 0.881 | 0.581 | 0.327 | 0.387 | 0.531 | 0.533 | 0.452 | 0.502 | 0.738 | 0.442 | 0.881 | 0.369 | 0.406 | 0.658 | 0.990 | 0.290 | 0.267 | 0.137 | 0.862 | 0.302 | 0.125 | 0.192 | 0.087 | 0.200 | 0.175 | 0.196 | 0.048 | 0.117 | 0.090 | 0.206 | 0.244 | 0.085 | 0.090 | 0.106 | 0.033 | 0.000 | 0.021 | 0.000 |
| LOC_STREET | ADDRESS | 0.980 | 0.975 | 0.964 | 0.957 | 0.987 | 0.954 | 0.973 | 0.925 | 0.931 | 0.931 | 0.918 | 0.829 | 0.764 | 0.944 | 0.946 | 0.971 | 0.882 | 0.961 | 0.928 | 0.916 | 0.869 | 0.918 | 0.925 | 0.914 | 0.923 | 0.888 | 0.980 | 0.961 | 0.728 | 0.843 | 0.898 | 0.889 | 0.650 | 0.632 | 0.614 | 0.857 | 0.450 | 0.661 | 0.721 | 0.467 | 0.688 | 0.648 | 0.644 | 0.494 | 0.562 | 0.490 | 0.226 | 0.205 | 0.148 | 0.102 | 0.115 | 0.022 | 0.000 | 0.000 | 0.000 |
| LOC_ZIP | ADDRESS | 0.854 | 0.868 | 0.875 | 0.847 | 0.924 | 0.840 | 0.854 | 0.882 | 0.882 | 0.382 | 0.757 | 0.833 | 0.646 | 0.771 | 0.896 | 0.896 | 0.208 | 0.569 | 0.868 | 0.215 | 0.194 | 0.972 | 0.958 | 0.972 | 0.868 | 0.847 | 0.229 | 0.264 | 0.764 | 0.938 | 0.562 | 0.611 | 0.132 | 0.972 | 0.646 | 0.361 | 0.729 | 0.889 | 0.896 | 0.764 | 0.889 | 0.840 | 0.958 | 0.792 | 0.812 | 0.812 | 0.347 | 0.403 | 0.167 | 0.340 | 0.410 | 0.111 | 0.000 | 0.000 | 0.000 |
| ORG | ORG | 0.904 | 0.823 | 0.854 | 0.842 | 0.820 | 0.862 | 0.803 | 0.782 | 0.777 | 0.859 | 0.867 | 0.808 | 0.790 | 0.396 | 0.404 | 0.419 | 0.659 | 0.833 | 0.380 | 0.582 | 0.606 | 0.195 | 0.196 | 0.196 | 0.331 | 0.573 | 0.560 | 0.772 | 0.245 | 0.237 | 0.641 | 0.922 | 0.243 | 0.244 | 0.029 | 0.584 | 0.068 | 0.070 | 0.106 | 0.073 | 0.080 | 0.103 | 0.123 | 0.044 | 0.054 | 0.043 | 0.188 | 0.156 | 0.040 | 0.053 | 0.046 | 0.043 | 0.000 | 0.013 | 0.000 |
| PERSON | PERSON | 0.927 | 0.864 | 0.912 | 0.940 | 0.851 | 0.940 | 0.888 | 0.817 | 0.821 | 0.904 | 0.903 | 0.845 | 0.837 | 0.853 | 0.842 | 0.836 | 0.899 | 0.871 | 0.843 | 0.908 | 0.902 | 0.953 | 0.936 | 0.957 | 0.837 | 0.901 | 0.793 | 0.849 | 0.828 | 0.781 | 0.710 | 0.666 | 0.832 | 0.761 | 0.913 | 0.551 | 0.830 | 0.818 | 0.777 | 0.764 | 0.615 | 0.642 | 0.430 | 0.561 | 0.520 | 0.537 | 0.199 | 0.260 | 0.144 | 0.114 | 0.234 | 0.114 | 0.000 | 0.025 | 0.002 |
| PERSON_EMAIL | CONTACT | 0.721 | 0.729 | 0.625 | 0.850 | 0.517 | 0.162 | 0.459 | 0.740 | 0.737 | 0.559 | 0.742 | 0.459 | 0.430 | 0.454 | 0.273 | 0.340 | 0.635 | 0.067 | 0.348 | 0.682 | 0.643 | 0.836 | 0.804 | 0.849 | 0.344 | 0.331 | 0.457 | 0.062 | 0.413 | 0.584 | 0.598 | 0.224 | 0.096 | 0.392 | 0.373 | 0.485 | 0.589 | 0.691 | 0.727 | 0.476 | 0.720 | 0.688 | 0.205 | 0.442 | 0.477 | 0.460 | 0.636 | 0.428 | 0.334 | 0.296 | 0.047 | 0.001 | 0.385 | 0.029 | 0.000 |

## char F1 by domain

| domain | n | gliner2-fastino | gliner-nvidia+sent300 | gliner2-large | nuner-zero | gliner2-hivetrace-omni | stanza-ru | bardsai-eu | gliner-nvidia+ov100 | gliner-nvidia | gliner-multi-v21 | gliner-urchade | gliner-pii-base | gliner-pii-edge | openmed-multilingual | nym-small | nym-base+sent300 | gliner25-fastino+ov100 | davlan-xlmr | nym-base | gliner25-fastino+sent300 | gliner25-fastino | pplx | pplx+sent300 | pplx+ov100 | nym-base+ov100 | gravitee-small | gliner2-hivetrace-uni | davlan-mbert | openmed-nemotron | kalyan-ettin | gliner25-fastino+nochunk | ner-ru-yqelz | ner-ru-gherman | mmbert32k | gliner2-vladlinv | gliner-stream-pii | ru-pii-ner | openai-base | apararti | opf-ru | opf-kz-ru | traciora | mmbert32k+nochunk | opf-ru-v2+ov100 | opf-ru-v2+sent300 | opf-ru-v2 | pii-shield-onnx | ru-legal-ner+sent300 | ru-legal-ner+ov100 | ru-legal-ner | spacy-ru-lg | natasha | rules-ru | fef2-secret-ru | spacy-alrosait |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| aidashboard | 20 | 0.532 | 0.547 | 0.530 | 0.464 | 0.705 | 0.131 | 0.861 | 0.638 | 0.570 | 0.501 | 0.475 | 0.666 | 0.769 | 0.568 | 0.794 | 0.810 | 0.736 | 0.849 | 0.800 | 0.624 | 0.724 | 0.835 | 0.827 | 0.867 | 0.794 | 0.830 | 0.781 | 0.823 | 0.466 | 0.610 | 0.851 | 0.698 | 0.778 | 0.532 | 0.802 | 0.681 | 0.561 | 0.439 | 0.425 | 0.402 | 0.282 | 0.486 | 0.437 | 0.359 | 0.379 | 0.388 | 0.273 | 0.228 | 0.156 | 0.127 | 0.207 | 0.177 | 0.000 | 0.052 | 0.020 |
| anamnesis | 20 | 0.561 | 0.517 | 0.554 | 0.492 | 0.673 | 0.173 | 0.737 | 0.550 | 0.532 | 0.488 | 0.468 | 0.605 | 0.726 | 0.603 | 0.831 | 0.812 | 0.652 | 0.866 | 0.815 | 0.603 | 0.633 | 0.697 | 0.716 | 0.686 | 0.813 | 0.708 | 0.745 | 0.845 | 0.542 | 0.581 | 0.859 | 0.623 | 0.740 | 0.580 | 0.710 | 0.718 | 0.438 | 0.390 | 0.412 | 0.377 | 0.392 | 0.654 | 0.629 | 0.578 | 0.576 | 0.610 | 0.290 | 0.322 | 0.299 | 0.278 | 0.258 | 0.153 | 0.005 | 0.046 | 0.000 |
| billing | 20 | 0.663 | 0.633 | 0.644 | 0.582 | 0.757 | 0.232 | 0.830 | 0.669 | 0.659 | 0.580 | 0.563 | 0.702 | 0.733 | 0.590 | 0.707 | 0.683 | 0.748 | 0.839 | 0.687 | 0.678 | 0.730 | 0.647 | 0.681 | 0.645 | 0.663 | 0.732 | 0.731 | 0.817 | 0.523 | 0.614 | 0.700 | 0.701 | 0.642 | 0.539 | 0.545 | 0.702 | 0.435 | 0.432 | 0.488 | 0.427 | 0.468 | 0.572 | 0.391 | 0.506 | 0.504 | 0.491 | 0.261 | 0.351 | 0.210 | 0.207 | 0.181 | 0.133 | 0.053 | 0.007 | 0.000 |
| carinsurance | 20 | 0.659 | 0.645 | 0.543 | 0.618 | 0.681 | 0.220 | 0.719 | 0.683 | 0.647 | 0.544 | 0.519 | 0.622 | 0.668 | 0.626 | 0.620 | 0.642 | 0.681 | 0.491 | 0.639 | 0.667 | 0.669 | 0.821 | 0.804 | 0.818 | 0.637 | 0.561 | 0.667 | 0.466 | 0.538 | 0.618 | 0.693 | 0.514 | 0.393 | 0.559 | 0.595 | 0.553 | 0.518 | 0.534 | 0.606 | 0.495 | 0.635 | 0.676 | 0.453 | 0.568 | 0.579 | 0.563 | 0.442 | 0.462 | 0.385 | 0.360 | 0.117 | 0.063 | 0.260 | 0.032 | 0.000 |
| grouptherapy | 7 | 0.478 | 0.523 | 0.512 | 0.396 | 0.693 | 0.128 | 0.867 | 0.585 | 0.572 | 0.446 | 0.451 | 0.550 | 0.630 | 0.561 | 0.928 | 0.933 | 0.715 | 0.944 | 0.946 | 0.507 | 0.663 | 0.929 | 0.926 | 0.926 | 0.929 | 0.775 | 0.852 | 0.919 | 0.543 | 0.757 | 0.897 | 0.717 | 0.873 | 0.565 | 0.806 | 0.795 | 0.628 | 0.363 | 0.342 | 0.360 | 0.259 | 0.468 | 0.667 | 0.281 | 0.235 | 0.243 | 0.387 | 0.226 | 0.101 | 0.048 | 0.141 | 0.165 | 0.000 | 0.015 | 0.000 |
| police | 20 | 0.586 | 0.613 | 0.509 | 0.523 | 0.599 | 0.198 | 0.753 | 0.670 | 0.627 | 0.528 | 0.491 | 0.576 | 0.690 | 0.601 | 0.707 | 0.728 | 0.613 | 0.650 | 0.677 | 0.570 | 0.595 | 0.792 | 0.780 | 0.784 | 0.663 | 0.627 | 0.672 | 0.622 | 0.520 | 0.631 | 0.680 | 0.626 | 0.503 | 0.502 | 0.573 | 0.585 | 0.514 | 0.435 | 0.496 | 0.377 | 0.514 | 0.573 | 0.479 | 0.481 | 0.480 | 0.494 | 0.381 | 0.411 | 0.368 | 0.289 | 0.139 | 0.093 | 0.163 | 0.093 | 0.000 |
| therapy | 20 | 0.663 | 0.630 | 0.688 | 0.568 | 0.805 | 0.195 | 0.897 | 0.685 | 0.651 | 0.584 | 0.591 | 0.704 | 0.825 | 0.588 | 0.865 | 0.849 | 0.712 | 0.917 | 0.861 | 0.593 | 0.682 | 0.806 | 0.759 | 0.784 | 0.846 | 0.789 | 0.782 | 0.911 | 0.512 | 0.638 | 0.890 | 0.849 | 0.729 | 0.512 | 0.634 | 0.835 | 0.419 | 0.230 | 0.253 | 0.254 | 0.213 | 0.276 | 0.411 | 0.129 | 0.146 | 0.136 | 0.304 | 0.234 | 0.061 | 0.094 | 0.143 | 0.135 | 0.000 | 0.017 | 0.000 |
| triage | 20 | 0.466 | 0.429 | 0.453 | 0.416 | 0.515 | 0.153 | 0.910 | 0.466 | 0.460 | 0.431 | 0.345 | 0.641 | 0.716 | 0.747 | 0.920 | 0.931 | 0.445 | 0.901 | 0.928 | 0.440 | 0.436 | 0.852 | 0.849 | 0.835 | 0.931 | 0.856 | 0.800 | 0.891 | 0.525 | 0.539 | 0.627 | 0.678 | 0.777 | 0.547 | 0.708 | 0.772 | 0.409 | 0.333 | 0.373 | 0.291 | 0.444 | 0.746 | 0.673 | 0.587 | 0.617 | 0.574 | 0.137 | 0.321 | 0.350 | 0.334 | 0.341 | 0.186 | 0.000 | 0.059 | 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 |
|---|---|---|---|---|
| gliner2-fastino | 313 (10.1%) | 211 (6.8%) | 169 (5.5%) | 104 (3.4%) |
| gliner-nvidia+sent300 | 410 (13.3%) | 313 (10.1%) | 282 (9.1%) | 225 (7.3%) |
| gliner2-large | 437 (14.2%) | 296 (9.6%) | 219 (7.1%) | 150 (4.9%) |
| nuner-zero | 452 (14.6%) | 292 (9.5%) | 209 (6.8%) | 135 (4.4%) |
| gliner2-hivetrace-omni | 518 (16.8%) | 370 (12.0%) | 296 (9.6%) | 219 (7.1%) |
| bardsai-eu | 527 (17.1%) | 492 (15.9%) | 489 (15.8%) | 489 (15.8%) |
| gliner-nvidia+ov100 | 531 (17.2%) | 407 (13.2%) | 348 (11.3%) | 284 (9.2%) |
| gliner-nvidia | 538 (17.4%) | 404 (13.1%) | 358 (11.6%) | 277 (9.0%) |
| gliner-multi-v21 | 544 (17.6%) | 273 (8.8%) | 197 (6.4%) | 115 (3.7%) |
| gliner-urchade | 570 (18.5%) | 433 (14.0%) | 346 (11.2%) | 262 (8.5%) |
| gliner-pii-base | 617 (20.0%) | 212 (6.9%) | 90 (2.9%) | 26 (0.8%) |
| gliner-pii-edge | 624 (20.2%) | 220 (7.1%) | 99 (3.2%) | 21 (0.7%) |
| openmed-multilingual | 660 (21.4%) | 572 (18.5%) | 568 (18.4%) | 568 (18.4%) |
| nym-small | 680 (22.0%) | 655 (21.2%) | 655 (21.2%) | 655 (21.2%) |
| nym-base+sent300 | 695 (22.5%) | 670 (21.7%) | 670 (21.7%) | 670 (21.7%) |
| gliner25-fastino+ov100 | 702 (22.7%) | 565 (18.3%) | 497 (16.1%) | 411 (13.3%) |
| davlan-xlmr | 718 (23.3%) | 711 (23.0%) | 711 (23.0%) | 711 (23.0%) |
| nym-base | 719 (23.3%) | 695 (22.5%) | 694 (22.5%) | 694 (22.5%) |
| gliner25-fastino+sent300 | 732 (23.7%) | 628 (20.3%) | 590 (19.1%) | 533 (17.3%) |
| gliner25-fastino | 742 (24.0%) | 616 (20.0%) | 543 (17.6%) | 457 (14.8%) |
| nym-base+ov100 | 754 (24.4%) | 733 (23.7%) | 733 (23.7%) | 733 (23.7%) |
| gravitee-small | 796 (25.8%) | 719 (23.3%) | 719 (23.3%) | 719 (23.3%) |
| gliner2-hivetrace-uni | 817 (26.5%) | 484 (15.7%) | 355 (11.5%) | 231 (7.5%) |
| davlan-mbert | 836 (27.1%) | 834 (27.0%) | 834 (27.0%) | 834 (27.0%) |
| openmed-nemotron | 906 (29.3%) | 804 (26.0%) | 797 (25.8%) | 797 (25.8%) |
| kalyan-ettin | 966 (31.3%) | 822 (26.6%) | 810 (26.2%) | 810 (26.2%) |
| gliner25-fastino+nochunk | 996 (32.3%) | 887 (28.7%) | 839 (27.2%) | 814 (26.4%) |
| ner-ru-yqelz | 1026 (33.2%) | 863 (28.0%) | 859 (27.8%) | 859 (27.8%) |
| ner-ru-gherman | 1053 (34.1%) | 972 (31.5%) | 958 (31.0%) | 957 (31.0%) |
| mmbert32k | 1095 (35.5%) | 753 (24.4%) | 710 (23.0%) | 706 (22.9%) |
| gliner2-vladlinv | 1179 (38.2%) | 1142 (37.0%) | 1126 (36.5%) | 1090 (35.3%) |
| gliner-stream-pii | 1233 (39.9%) | 963 (31.2%) | 843 (27.3%) | 654 (21.2%) |
| openai-base | 1309 (42.4%) | 1308 (42.4%) | 1308 (42.4%) | 1308 (42.4%) |
| apararti | 1334 (43.2%) | 1326 (43.0%) | 1325 (42.9%) | 1325 (42.9%) |
| opf-ru | 1349 (43.7%) | 1326 (43.0%) | 1325 (42.9%) | 1325 (42.9%) |
| opf-kz-ru | 1594 (51.6%) | 1586 (51.4%) | 1586 (51.4%) | 1586 (51.4%) |
| traciora | 1655 (53.6%) | 1640 (53.1%) | 1639 (53.1%) | 1638 (53.1%) |
| mmbert32k+nochunk | 1805 (58.5%) | 1378 (44.6%) | 1316 (42.6%) | 1311 (42.5%) |
| opf-ru-v2+ov100 | 1904 (61.7%) | 1897 (61.5%) | 1897 (61.5%) | 1897 (61.5%) |
| opf-ru-v2+sent300 | 1925 (62.4%) | 1917 (62.1%) | 1917 (62.1%) | 1917 (62.1%) |
| opf-ru-v2 | 1936 (62.7%) | 1930 (62.5%) | 1930 (62.5%) | 1930 (62.5%) |
| pii-shield-onnx | 2242 (72.6%) | 1966 (63.7%) | 1943 (62.9%) | 1942 (62.9%) |
| ru-legal-ner+sent300 | 2244 (72.7%) | 1494 (48.4%) | 1405 (45.5%) | 1389 (45.0%) |
| ru-legal-ner+ov100 | 2583 (83.7%) | 1983 (64.2%) | 1848 (59.9%) | 1834 (59.4%) |
| ru-legal-ner | 2632 (85.3%) | 2078 (67.3%) | 1964 (63.6%) | 1939 (62.8%) |
| fef2-secret-ru | 3032 (98.2%) | 3030 (98.2%) | 3030 (98.2%) | 3030 (98.2%) |

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.