# synth-wiki-tables - ru / pii (300 rows, 13094 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 0.606 (39.4% 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 | **489** | 3.7% [3.4%, 4.1%] | 74.5% | 0.932 [0.929, 0.935] | 0.976 | 0.892 | 0.632 | 0.958 | 0.620 | 0/0 | 0 | 0 | 392 | - |  |
| pplx | **734** | 5.6% [4.8%, 6.5%] | 88.9% | 0.944 [0.938, 0.950] | 0.993 | 0.900 | 0.909 | 0.965 | 0.714 | 0/0 | 0 | 0 | 237 | 596M |  |
| ru-legal-ner | **776** | 5.9% [5.3%, 6.6%] | 63.5% | 0.815 [0.797, 0.832] | 0.882 | 0.757 | 0.227 | 0.929 | 0.517 | 0/0 | 0 | 0 | 20 | 29M |  |
| nym-base | **897** | 6.9% [5.6%, 8.2%] | 68.6% | 0.842 [0.821, 0.864] | 0.994 | 0.730 | 0.608 | 0.958 | 0.488 | 0/0 | 0 | 0 | 22 | 308M |  |
| apararti | **1221** | 9.3% [8.2%, 10.5%] | 82.0% | 0.905 [0.894, 0.917] | 0.978 | 0.842 | 0.757 | 0.932 | 0.640 | 0/0 | 0 | 0 | 135 | 1.4B |  |
| gliner2-fastino-ru | **1259** | 9.6% [8.1%, 11.2%] | 85.3% | 0.910 [0.901, 0.919] | 0.990 | 0.842 | 0.885 | 0.942 | 0.765 | 0/0 | 0 | 45 | 39 | 307M |  |
| opf-kz-ru | **1494** | 11.4% [10.6%, 12.3%] | 79.4% | 0.900 [0.892, 0.909] | 0.969 | 0.841 | 0.753 | 0.923 | 0.652 | 0/0 | 0 | 0 | 167 | 1.4B |  |
| gliner2-fastino | **1549** | 11.8% [10.1%, 13.5%] | 85.7% | 0.898 [0.886, 0.910] | 0.995 | 0.818 | 0.900 | 0.933 | 0.698 | 0/0 | 0 | 50 | 25 | 307M |  |
| openmed-nemotron | **1675** | 12.8% [11.8%, 13.8%] | 65.5% | 0.865 [0.859, 0.870] | 0.906 | 0.827 | 0.579 | 0.873 | 0.719 | 0/0 | 0 | 0 | 82 | 1.4B |  |
| openai-base | **1698** | 13.0% [11.4%, 14.7%] | 80.9% | 0.881 [0.866, 0.897] | 0.980 | 0.800 | 0.820 | 0.920 | 0.634 | 0/0 | 0 | 0 | 200 | 1.4B |  |
| mmbert32k | **1703** | 13.0% [12.5%, 13.6%] | 28.5% | 0.737 [0.713, 0.759] | 0.983 | 0.589 | 0.220 | 0.911 | 0.457 | 0/0 | 0 | 0 | 31 | 308M |  |
| openmed-multilingual | **2016** | 15.4% [14.7%, 16.1%] | 61.8% | 0.849 [0.846, 0.853] | 0.951 | 0.767 | 0.605 | 0.883 | 0.704 | 0/0 | 0 | 0 | 86 | 1.4B |  |
| bardsai-eu | **2378** | 18.2% [16.3%, 20.3%] | 45.6% | 0.705 [0.683, 0.726] | 0.795 | 0.634 | 0.414 | 0.834 | 0.207 | 0/0 | 0 | 0 | 1405 | - |  |
| gliner2-hivetrace-omni-ru | **2402** | 18.3% [16.9%, 19.8%] | 80.9% | 0.888 [0.880, 0.897] | 0.993 | 0.804 | 0.886 | 0.894 | 0.705 | 0/0 | 0 | 0 | 35 | 307M |  |
| gliner2-hivetrace-omni | **2632** | 20.1% [18.0%, 22.2%] | 79.6% | 0.843 [0.825, 0.861] | 0.989 | 0.734 | 0.877 | 0.881 | 0.682 | 0/0 | 0 | 0 | 58 | 307M |  |
| nuner-zero | **2767** | 21.1% [20.1%, 22.2%] | 64.0% | 0.903 [0.898, 0.908] | 0.979 | 0.838 | 0.686 | 0.863 | 0.880 | 0/0 | 0 | 0 | 29 | 449M |  |
| gliner-pii-edge | **2951** | 22.5% [20.3%, 24.8%] | 55.2% | 0.773 [0.749, 0.798] | 0.968 | 0.643 | 0.600 | 0.850 | 0.468 | 0/0 | 0 | 0 | 660 | 45M |  |
| traciora | **2970** | 22.7% [19.9%, 25.8%] | 65.3% | 0.765 [0.732, 0.796] | 0.995 | 0.622 | 0.684 | 0.868 | 0.522 | 0/0 | 0 | 0 | 711 | 1.4B |  |
| pii-shield-onnx | **3002** | 22.9% [21.6%, 24.4%] | 63.4% | 0.878 [0.872, 0.883] | 0.938 | 0.824 | 0.427 | 0.846 | 0.336 | 0/0 | 0 | 0 | 507 | - |  |
| gliner2-large | **3202** | 24.5% [23.5%, 25.4%] | 74.2% | 0.873 [0.866, 0.880] | 0.978 | 0.788 | 0.818 | 0.851 | 0.798 | 0/0 | 0 | 0 | 72 | 486M |  |
| kalyan-ettin | **3207** | 24.5% [23.5%, 25.6%] | 33.5% | 0.729 [0.708, 0.751] | 0.980 | 0.581 | 0.329 | 0.843 | 0.630 | 0/0 | 0 | 0 | 51 | 68M |  |
| opf-ru | **3226** | 24.6% [22.3%, 27.1%] | 49.8% | 0.742 [0.710, 0.773] | 0.990 | 0.593 | 0.525 | 0.850 | 0.574 | 0/0 | 0 | 0 | 97 | 1.4B |  |
| gliner2-vladlinv-ru | **3449** | 26.3% [22.0%, 30.8%] | 69.1% | 0.744 [0.703, 0.784] | 0.999 | 0.593 | 0.795 | 0.848 | 0.518 | 0/0 | 0 | 0 | 21 | 287M |  |
| opf-ru-v2 | **3468** | 26.5% [23.9%, 29.3%] | 63.3% | 0.738 [0.706, 0.769] | 0.988 | 0.589 | 0.693 | 0.834 | 0.534 | 0/0 | 0 | 0 | 1492 | 1.4B |  |
| gliner2-vladlinv | **3476** | 26.5% [22.2%, 31.0%] | 69.1% | 0.744 [0.703, 0.784] | 0.999 | 0.592 | 0.797 | 0.847 | 0.515 | 0/0 | 0 | 0 | 13 | 287M |  |
| ru-pii-ner | **3765** | 28.8% [24.5%, 33.0%] | 66.7% | 0.725 [0.685, 0.765] | 0.988 | 0.573 | 0.769 | 0.826 | 0.372 | 0/0 | 0 | 0 | 308 | 358M |  |
| rules-ru | **3967** | 30.3% [28.4%, 32.1%] | 69.7% | 0.867 [0.854, 0.880] | 0.999 | 0.766 | 0.803 | 0.821 | 0.864 | 0/0 | 0 | 0 | 2 | - |  |
| gliner25-fastino-ru | **3988** | 30.5% [28.6%, 32.4%] | 68.2% | 0.794 [0.774, 0.812] | 0.999 | 0.658 | 0.804 | 0.820 | 0.671 | 0/0 | 0 | 0 | 24 | 287M |  |
| gliner25-fastino | **4548** | 34.7% [32.4%, 37.1%] | 63.8% | 0.768 [0.744, 0.791] | 0.999 | 0.623 | 0.769 | 0.790 | 0.648 | 0/0 | 0 | 0 | 12 | 287M |  |
| gliner-urchade | **4657** | 35.6% [32.8%, 38.3%] | 62.2% | 0.709 [0.678, 0.738] | 0.993 | 0.551 | 0.749 | 0.781 | 0.536 | 0/0 | 0 | 0 | 25 | 289M |  |
| gliner-nvidia | **5022** | 38.4% [37.5%, 39.1%] | 53.5% | 0.708 [0.698, 0.717] | 0.969 | 0.558 | 0.646 | 0.747 | 0.679 | 0/0 | 0 | 0 | 39 | 445M |  |
| gliner-stream-pii | **5617** | 42.9% [41.5%, 44.2%] | 53.7% | 0.637 [0.616, 0.658] | 0.966 | 0.475 | 0.667 | 0.720 | 0.533 | 0/0 | 0 | 0 | 23 | 677M |  |
| gliner-urchade-ru | **5664** | 43.3% [40.8%, 45.8%] | 52.4% | 0.614 [0.585, 0.642] | 0.996 | 0.443 | 0.662 | 0.724 | 0.316 | 0/0 | 0 | 0 | 29 | 289M |  |
| gliner-nvidia-ru | **6820** | 52.1% [51.3%, 52.8%] | 41.1% | 0.600 [0.584, 0.616] | 0.980 | 0.433 | 0.548 | 0.641 | 0.549 | 0/0 | 0 | 0 | 131 | 445M |  |
| gravitee-small | **7163** | 54.7% [54.0%, 55.4%] | 37.0% | 0.579 [0.558, 0.599] | 0.831 | 0.445 | 0.459 | 0.580 | 0.556 | 0/0 | 0 | 0 | 25 | 29M |  |
| gliner2-hivetrace-uni | **7628** | 58.3% [55.9%, 60.7%] | 40.8% | 0.580 [0.544, 0.615] | 0.967 | 0.414 | 0.565 | 0.578 | 0.555 | 0/0 | 0 | 0 | 33 | 147M |  |
| fef2-secret-ru | **7959** | 60.8% [58.6%, 63.0%] | 24.8% | 0.551 [0.518, 0.583] | 0.983 | 0.382 | 0.275 | 0.558 | 0.123 | 0/0 | 0 | 0 | 22 | 177M |  |
| ner-ru-gherman | **7972** | 60.9% [58.7%, 63.3%] | 3.9% | 0.421 [0.392, 0.448] | 1.000 | 0.267 | 0.049 | 0.562 | 0.000 | 0/0 | 0 | 0 | 22 | 177M |  |
| gliner-pii-base | **8777** | 67.0% [66.3%, 67.7%] | 31.9% | 0.433 [0.414, 0.450] | 0.980 | 0.278 | 0.474 | 0.491 | 0.432 | 0/0 | 0 | 0 | 16 | 166M |  |
| gliner-multi-v21 | **10169** | 77.7% [76.5%, 79.0%] | 18.1% | 0.317 [0.298, 0.335] | 0.927 | 0.191 | 0.281 | 0.351 | 0.263 | 0/0 | 0 | 0 | 17 | 289M |  |
| davlan-xlmr | **10468** | 79.9% [78.8%, 81.1%] | 14.8% | 0.311 [0.291, 0.332] | 0.989 | 0.185 | 0.242 | 0.334 | 0.259 | 0/0 | 0 | 0 | 20 | 277M |  |
| ner-ru-yqelz | **10705** | 81.8% [81.4%, 82.1%] | 14.4% | 0.297 [0.283, 0.309] | 0.929 | 0.176 | 0.231 | 0.299 | 0.263 | 0/0 | 0 | 0 | 23 | 559M |  |
| stanza-ru | **11005** | 84.0% [83.2%, 84.8%] | 13.7% | 0.264 [0.246, 0.282] | 0.905 | 0.155 | 0.220 | 0.258 | 0.161 | 0/0 | 0 | 0 | 204 | - |  |
| davlan-mbert | **11209** | 85.6% [84.9%, 86.3%] | 13.6% | 0.266 [0.248, 0.284] | 0.995 | 0.153 | 0.236 | 0.252 | 0.262 | 0/0 | 0 | 0 | 20 | 177M |  |
| gliner-multi-v21-ru | **11271** | 86.1% [85.3%, 86.8%] | 12.0% | 0.206 [0.192, 0.219] | 0.905 | 0.116 | 0.199 | 0.236 | 0.182 | 0/0 | 0 | 0 | 30 | 289M |  |
| spacy-alrosait | **11480** | 87.7% [87.0%, 88.4%] | 11.8% | 0.236 [0.218, 0.254] | 1.000 | 0.134 | 0.209 | 0.219 | 0.000 | 0/0 | 0 | 0 | 19 | - |  |
| natasha | **11506** | 87.9% [87.3%, 88.5%] | 9.8% | 0.218 [0.203, 0.233] | 0.944 | 0.123 | 0.163 | 0.215 | 0.215 | 0/0 | 0 | 0 | 11 | - |  |
| spacy-ru-lg | **12011** | 91.7% [91.3%, 92.2%] | 3.4% | 0.138 [0.130, 0.147] | 0.723 | 0.077 | 0.043 | 0.151 | 0.124 | 0/0 | 0 | 0 | 20 | - |  |
| gliner2-hivetrace-uni-ru | **12306** | 94.0% [93.4%, 94.5%] | 5.7% | 0.081 [0.073, 0.090] | 0.823 | 0.043 | 0.106 | 0.111 | 0.045 | 0/0 | 0 | 0 | 39 | 147M |  |

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`: pplx ≈ ru-legal-ner; apararti ≈ gliner2-fastino-ru; opf-kz-ru ≈ gliner2-fastino; gliner2-fastino ≈ openmed-nemotron; openmed-nemotron ≈ openai-base; openai-base ≈ mmbert32k; bardsai-eu ≈ gliner2-hivetrace-omni-ru; gliner2-hivetrace-omni ≈ nuner-zero; nuner-zero ≈ gliner-pii-edge; gliner-pii-edge ≈ traciora; traciora ≈ pii-shield-onnx; pii-shield-onnx ≈ gliner2-large; gliner2-large ≈ kalyan-ettin; kalyan-ettin ≈ opf-ru; opf-ru ≈ gliner2-vladlinv-ru; gliner2-vladlinv-ru ≈ opf-ru-v2; opf-ru-v2 ≈ gliner2-vladlinv; ru-pii-ner ≈ rules-ru; rules-ru ≈ gliner25-fastino-ru; gliner25-fastino ≈ gliner-urchade; gliner-urchade ≈ gliner-nvidia; gliner-stream-pii ≈ gliner-urchade-ru; fef2-secret-ru ≈ ner-ru-gherman; davlan-mbert ≈ gliner-multi-v21-ru; spacy-alrosait ≈ natasha

## Missed by group

| model | PERSON | CONTACT | NET | ACCOUNT | ORG |
|---|---|---|---|---|---|
| spans in gold | 1597 | 4791 | 3212 | 3194 | 300 |
| nym-small | 0 (0.0%) | 99 (2.1%) | 109 (3.4%) | 278 (8.7%) | 3 (1.0%) |
| pplx | 0 (0.0%) | 0 (0.0%) | 532 (16.6%) | 1 (0.0%) | 201 (67.0%) |
| ru-legal-ner | 11 (0.7%) | 7 (0.1%) | 427 (13.3%) | 283 (8.9%) | 48 (16.0%) |
| nym-base | 0 (0.0%) | 12 (0.3%) | 869 (27.1%) | 16 (0.5%) | 0 (0.0%) |
| apararti | 15 (0.9%) | 4 (0.1%) | 802 (25.0%) | 292 (9.1%) | 108 (36.0%) |
| gliner2-fastino-ru | 8 (0.5%) | 0 (0.0%) | 1202 (37.4%) | 1 (0.0%) | 48 (16.0%) |
| opf-kz-ru | 145 (9.1%) | 2 (0.0%) | 642 (20.0%) | 495 (15.5%) | 210 (70.0%) |
| gliner2-fastino | 2 (0.1%) | 0 (0.0%) | 1340 (41.7%) | 170 (5.3%) | 37 (12.3%) |
| openmed-nemotron | 20 (1.3%) | 143 (3.0%) | 767 (23.9%) | 596 (18.7%) | 149 (49.7%) |
| openai-base | 26 (1.6%) | 2 (0.0%) | 1158 (36.1%) | 343 (10.7%) | 169 (56.3%) |
| mmbert32k | 18 (1.1%) | 381 (8.0%) | 445 (13.9%) | 800 (25.0%) | 59 (19.7%) |
| openmed-multilingual | 584 (36.6%) | 69 (1.4%) | 685 (21.3%) | 449 (14.1%) | 229 (76.3%) |
| bardsai-eu | 0 (0.0%) | 41 (0.9%) | 1529 (47.6%) | 773 (24.2%) | 35 (11.7%) |
| gliner2-hivetrace-omni-ru | 2 (0.1%) | 42 (0.9%) | 1451 (45.2%) | 863 (27.0%) | 44 (14.7%) |
| gliner2-hivetrace-omni | 0 (0.0%) | 60 (1.3%) | 1810 (56.4%) | 733 (22.9%) | 29 (9.7%) |
| nuner-zero | 1 (0.1%) | 1164 (24.3%) | 192 (6.0%) | 1285 (40.2%) | 125 (41.7%) |
| gliner-pii-edge | 303 (19.0%) | 193 (4.0%) | 1813 (56.4%) | 612 (19.2%) | 30 (10.0%) |
| traciora | 52 (3.3%) | 65 (1.4%) | 2316 (72.1%) | 500 (15.7%) | 37 (12.3%) |
| pii-shield-onnx | 651 (40.8%) | 353 (7.4%) | 14 (0.4%) | 1750 (54.8%) | 234 (78.0%) |
| gliner2-large | 52 (3.3%) | 674 (14.1%) | 977 (30.4%) | 1285 (40.2%) | 214 (71.3%) |
| kalyan-ettin | 27 (1.7%) | 545 (11.4%) | 1194 (37.2%) | 1328 (41.6%) | 113 (37.7%) |
| opf-ru | 4 (0.3%) | 23 (0.5%) | 2007 (62.5%) | 1122 (35.1%) | 70 (23.3%) |
| gliner2-vladlinv-ru | 1 (0.1%) | 0 (0.0%) | 3212 (100.0%) | 5 (0.2%) | 231 (77.0%) |
| opf-ru-v2 | 19 (1.2%) | 32 (0.7%) | 2330 (72.5%) | 970 (30.4%) | 117 (39.0%) |
| gliner2-vladlinv | 2 (0.1%) | 0 (0.0%) | 3212 (100.0%) | 36 (1.1%) | 226 (75.3%) |
| ru-pii-ner | 0 (0.0%) | 300 (6.3%) | 3181 (99.0%) | 81 (2.5%) | 203 (67.7%) |
| rules-ru | 1597 (100.0%) | 0 (0.0%) | 0 (0.0%) | 2300 (72.0%) | 70 (23.3%) |
| gliner25-fastino-ru | 1 (0.1%) | 145 (3.0%) | 2048 (63.8%) | 1665 (52.1%) | 129 (43.0%) |
| gliner25-fastino | 0 (0.0%) | 37 (0.8%) | 2341 (72.9%) | 2092 (65.5%) | 78 (26.0%) |
| gliner-urchade | 0 (0.0%) | 690 (14.4%) | 2624 (81.7%) | 1333 (41.7%) | 10 (3.3%) |
| gliner-nvidia | 293 (18.3%) | 1588 (33.1%) | 1301 (40.5%) | 1778 (55.7%) | 62 (20.7%) |
| gliner-stream-pii | 1092 (68.4%) | 648 (13.5%) | 1949 (60.7%) | 1790 (56.0%) | 138 (46.0%) |
| gliner-urchade-ru | 1596 (99.9%) | 496 (10.4%) | 2688 (83.7%) | 843 (26.4%) | 41 (13.7%) |
| gliner-nvidia-ru | 626 (39.2%) | 1748 (36.5%) | 1931 (60.1%) | 2387 (74.7%) | 128 (42.7%) |
| gravitee-small | 316 (19.8%) | 1702 (35.5%) | 1869 (58.2%) | 2981 (93.3%) | 295 (98.3%) |
| gliner2-hivetrace-uni | 162 (10.1%) | 1615 (33.7%) | 3198 (99.6%) | 2492 (78.0%) | 161 (53.7%) |
| fef2-secret-ru | 1 (0.1%) | 2444 (51.0%) | 3143 (97.9%) | 2334 (73.1%) | 37 (12.3%) |
| ner-ru-gherman | 0 (0.0%) | 2629 (54.9%) | 3212 (100.0%) | 2075 (65.0%) | 56 (18.7%) |
| gliner-pii-base | 1506 (94.3%) | 2588 (54.0%) | 2477 (77.1%) | 2041 (63.9%) | 165 (55.0%) |
| gliner-multi-v21 | 4 (0.3%) | 4264 (89.0%) | 2998 (93.3%) | 2765 (86.6%) | 138 (46.0%) |
| davlan-xlmr | 0 (0.0%) | 4264 (89.0%) | 3212 (100.0%) | 2992 (93.7%) | 0 (0.0%) |
| ner-ru-yqelz | 0 (0.0%) | 4737 (98.9%) | 2773 (86.3%) | 3194 (100.0%) | 1 (0.3%) |
| stanza-ru | 0 (0.0%) | 4790 (100.0%) | 3211 (100.0%) | 3004 (94.1%) | 0 (0.0%) |
| davlan-mbert | 1 (0.1%) | 4781 (99.8%) | 3212 (100.0%) | 3190 (99.9%) | 25 (8.3%) |
| gliner-multi-v21-ru | 609 (38.1%) | 4569 (95.4%) | 2955 (92.0%) | 2998 (93.9%) | 140 (46.7%) |
| spacy-alrosait | 39 (2.4%) | 4791 (100.0%) | 3212 (100.0%) | 3194 (100.0%) | 244 (81.3%) |
| natasha | 284 (17.8%) | 4791 (100.0%) | 3212 (100.0%) | 3194 (100.0%) | 25 (8.3%) |
| spacy-ru-lg | 815 (51.0%) | 4772 (99.6%) | 3163 (98.5%) | 3191 (99.9%) | 70 (23.3%) |
| gliner2-hivetrace-uni-ru | 1403 (87.9%) | 4565 (95.3%) | 3209 (99.9%) | 2863 (89.6%) | 266 (88.7%) |

## Char recall by gold type

| type | group | nym-small | pplx | ru-legal-ner | nym-base | apararti | gliner2-fastino-ru | opf-kz-ru | gliner2-fastino | openmed-nemotron | openai-base | mmbert32k | openmed-multilingual | bardsai-eu | gliner2-hivetrace-omni-ru | gliner2-hivetrace-omni | nuner-zero | gliner-pii-edge | traciora | pii-shield-onnx | gliner2-large | kalyan-ettin | opf-ru | gliner2-vladlinv-ru | opf-ru-v2 | gliner2-vladlinv | ru-pii-ner | rules-ru | gliner25-fastino-ru | gliner25-fastino | gliner-urchade | gliner-nvidia | gliner-stream-pii | gliner-urchade-ru | gliner-nvidia-ru | gravitee-small | gliner2-hivetrace-uni | fef2-secret-ru | ner-ru-gherman | gliner-pii-base | gliner-multi-v21 | davlan-xlmr | ner-ru-yqelz | stanza-ru | davlan-mbert | gliner-multi-v21-ru | spacy-alrosait | natasha | spacy-ru-lg | gliner2-hivetrace-uni-ru |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| email | CONTACT | 1.000 | 1.000 | 0.959 | 0.999 | 0.997 | 1.000 | 0.998 | 1.000 | 0.992 | 0.999 | 0.832 | 0.973 | 0.875 | 1.000 | 1.000 | 1.000 | 0.950 | 1.000 | 0.997 | 1.000 | 0.904 | 0.966 | 1.000 | 0.978 | 1.000 | 0.985 | 1.000 | 1.000 | 1.000 | 0.761 | 0.992 | 0.895 | 0.854 | 0.987 | 0.952 | 0.992 | 0.979 | 0.243 | 0.650 | 0.068 | 0.044 | 0.000 | 0.001 | 0.001 | 0.032 | 0.000 | 0.000 | 0.006 | 0.044 |
| employee_id | ACCOUNT | 0.766 | 0.766 | 0.726 | 0.762 | 0.818 | 0.777 | 0.846 | 0.793 | 0.472 | 0.803 | 0.645 | 0.694 | 0.441 | 0.892 | 0.762 | 0.811 | 0.645 | 0.549 | 0.213 | 0.814 | 0.117 | 0.226 | 0.765 | 0.598 | 0.755 | 0.761 | 0.000 | 0.198 | 0.051 | 0.345 | 0.318 | 0.464 | 0.490 | 0.175 | 0.035 | 0.091 | 0.041 | 0.000 | 0.505 | 0.036 | 0.000 | 0.000 | 0.036 | 0.000 | 0.014 | 0.000 | 0.000 | 0.001 | 0.106 |
| handle | CONTACT | 0.931 | 1.000 | 0.998 | 0.983 | 0.999 | 1.000 | 0.999 | 1.000 | 0.975 | 1.000 | 0.714 | 0.900 | 0.904 | 0.999 | 1.000 | 0.513 | 0.871 | 0.976 | 0.759 | 0.650 | 0.686 | 0.930 | 1.000 | 0.981 | 1.000 | 0.959 | 1.000 | 0.909 | 0.975 | 0.990 | 0.251 | 0.880 | 0.978 | 0.105 | 0.063 | 0.093 | 0.442 | 0.662 | 0.000 | 0.077 | 0.186 | 0.000 | 0.000 | 0.004 | 0.023 | 0.000 | 0.000 | 0.001 | 0.074 |
| hostname | NET | 0.892 | 0.281 | 0.417 | 0.056 | 0.030 | 0.899 | 0.164 | 0.852 | 0.001 | 0.091 | 0.053 | 0.042 | 0.065 | 0.353 | 0.369 | 0.999 | 0.301 | 0.007 | 0.978 | 1.000 | 0.000 | 0.009 | 0.000 | 0.000 | 0.000 | 0.018 | 1.000 | 0.175 | 0.284 | 0.232 | 0.788 | 0.037 | 0.153 | 0.027 | 0.000 | 0.005 | 0.000 | 0.000 | 0.001 | 0.006 | 0.000 | 0.177 | 0.000 | 0.000 | 0.002 | 0.000 | 0.000 | 0.052 | 0.001 |
| ip | NET | 0.981 | 0.996 | 0.463 | 0.901 | 0.957 | 1.000 | 0.981 | 1.000 | 0.962 | 0.877 | 0.847 | 0.995 | 0.480 | 0.996 | 0.996 | 1.000 | 0.617 | 0.740 | 0.906 | 0.965 | 0.817 | 0.176 | 0.000 | 0.968 | 0.000 | 0.017 | 1.000 | 0.887 | 0.465 | 0.550 | 0.770 | 0.898 | 0.541 | 0.686 | 0.959 | 0.000 | 0.000 | 0.000 | 0.972 | 0.243 | 0.000 | 0.002 | 0.001 | 0.000 | 0.333 | 0.000 | 0.000 | 0.008 | 0.000 |
| login | ACCOUNT | 0.800 | 1.000 | 0.991 | 0.994 | 0.836 | 1.000 | 0.767 | 0.997 | 0.910 | 0.956 | 0.476 | 0.823 | 0.931 | 0.521 | 0.734 | 0.285 | 0.820 | 0.877 | 0.642 | 0.340 | 0.944 | 0.910 | 0.998 | 0.498 | 0.997 | 0.963 | 0.561 | 0.728 | 0.630 | 0.763 | 0.517 | 0.324 | 0.865 | 0.262 | 0.067 | 0.323 | 0.464 | 0.623 | 0.131 | 0.208 | 0.111 | 0.000 | 0.000 | 0.002 | 0.101 | 0.000 | 0.000 | 0.001 | 0.081 |
| mac | NET | 1.000 | 1.000 | 0.299 | 1.000 | 0.927 | 0.000 | 0.828 | 0.000 | 0.876 | 0.606 | 0.488 | 0.995 | 0.248 | 0.000 | 0.000 | 0.669 | 0.027 | 0.010 | 0.991 | 0.000 | 0.751 | 0.083 | 0.000 | 0.005 | 0.000 | 0.003 | 1.000 | 0.009 | 0.082 | 0.000 | 0.989 | 0.578 | 0.000 | 1.000 | 0.415 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.004 | 0.000 |
| org | ORG | 0.988 | 0.160 | 0.842 | 0.998 | 0.478 | 0.689 | 0.191 | 0.674 | 0.310 | 0.324 | 0.503 | 0.113 | 0.682 | 0.864 | 0.879 | 0.539 | 0.471 | 0.597 | 0.133 | 0.119 | 0.323 | 0.481 | 0.091 | 0.389 | 0.104 | 0.165 | 0.803 | 0.331 | 0.444 | 0.918 | 0.661 | 0.622 | 0.809 | 0.457 | 0.007 | 0.291 | 0.749 | 0.457 | 0.347 | 0.348 | 0.929 | 0.983 | 0.772 | 0.807 | 0.331 | 0.113 | 0.838 | 0.676 | 0.079 |
| person | PERSON | 0.947 | 1.000 | 0.992 | 0.950 | 0.973 | 0.992 | 0.893 | 0.998 | 0.903 | 0.976 | 0.857 | 0.428 | 1.000 | 0.998 | 1.000 | 0.999 | 0.699 | 0.966 | 0.601 | 0.939 | 0.796 | 0.950 | 0.999 | 0.976 | 0.999 | 1.000 | 0.000 | 0.996 | 0.997 | 1.000 | 0.554 | 0.262 | 0.000 | 0.427 | 0.814 | 0.867 | 0.971 | 0.925 | 0.041 | 0.997 | 1.000 | 1.000 | 0.998 | 0.998 | 0.576 | 0.978 | 0.773 | 0.405 | 0.104 |
| phone | CONTACT | 1.000 | 1.000 | 0.997 | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | 0.941 | 1.000 | 0.890 | 0.988 | 0.995 | 0.981 | 0.972 | 0.719 | 0.998 | 0.985 | 0.999 | 0.950 | 0.841 | 1.000 | 1.000 | 1.000 | 1.000 | 0.838 | 1.000 | 1.000 | 1.000 | 0.807 | 0.805 | 0.832 | 0.861 | 0.835 | 0.894 | 0.927 | 0.001 | 0.000 | 0.779 | 0.055 | 0.000 | 0.008 | 0.000 | 0.000 | 0.032 | 0.000 | 0.000 | 0.002 | 0.016 |
| url | NET | 0.710 | 0.983 | 0.458 | 0.169 | 0.848 | 0.620 | 0.907 | 0.508 | 0.999 | 0.683 | 0.238 | 0.961 | 0.229 | 0.775 | 0.379 | 0.958 | 0.353 | 0.091 | 0.993 | 0.788 | 0.205 | 0.195 | 0.000 | 0.025 | 0.000 | 0.001 | 1.000 | 0.376 | 0.256 | 0.002 | 0.000 | 0.046 | 0.000 | 0.000 | 0.101 | 0.000 | 0.025 | 0.000 | 0.000 | 0.003 | 0.000 | 0.016 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |

## char F1 by domain

| domain | n | nym-small | pplx | ru-legal-ner | nym-base | apararti | gliner2-fastino-ru | opf-kz-ru | gliner2-fastino | openmed-nemotron | openai-base | mmbert32k | openmed-multilingual | bardsai-eu | gliner2-hivetrace-omni-ru | gliner2-hivetrace-omni | nuner-zero | gliner-pii-edge | traciora | pii-shield-onnx | gliner2-large | kalyan-ettin | opf-ru | gliner2-vladlinv-ru | opf-ru-v2 | gliner2-vladlinv | ru-pii-ner | rules-ru | gliner25-fastino-ru | gliner25-fastino | gliner-urchade | gliner-nvidia | gliner-stream-pii | gliner-urchade-ru | gliner-nvidia-ru | gravitee-small | gliner2-hivetrace-uni | fef2-secret-ru | ner-ru-gherman | gliner-pii-base | gliner-multi-v21 | davlan-xlmr | ner-ru-yqelz | stanza-ru | davlan-mbert | gliner-multi-v21-ru | spacy-alrosait | natasha | spacy-ru-lg | gliner2-hivetrace-uni-ru |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| table | 300 | 0.932 | 0.944 | 0.815 | 0.842 | 0.905 | 0.910 | 0.900 | 0.898 | 0.865 | 0.881 | 0.737 | 0.849 | 0.705 | 0.888 | 0.843 | 0.903 | 0.773 | 0.765 | 0.878 | 0.873 | 0.729 | 0.742 | 0.744 | 0.738 | 0.744 | 0.725 | 0.867 | 0.794 | 0.768 | 0.709 | 0.708 | 0.637 | 0.614 | 0.600 | 0.579 | 0.580 | 0.551 | 0.421 | 0.433 | 0.317 | 0.311 | 0.297 | 0.264 | 0.266 | 0.206 | 0.236 | 0.218 | 0.138 | 0.081 |

## 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 | 489 (3.7%) | 486 (3.7%) | 486 (3.7%) | 486 (3.7%) |
| ru-legal-ner | 776 (5.9%) | 264 (2.0%) | 239 (1.8%) | 239 (1.8%) |
| nym-base | 897 (6.9%) | 841 (6.4%) | 840 (6.4%) | 840 (6.4%) |
| apararti | 1221 (9.3%) | 1193 (9.1%) | 1191 (9.1%) | 1191 (9.1%) |
| gliner2-fastino-ru | 1259 (9.6%) | 966 (7.4%) | 887 (6.8%) | 792 (6.0%) |
| opf-kz-ru | 1494 (11.4%) | 1458 (11.1%) | 1456 (11.1%) | 1456 (11.1%) |
| gliner2-fastino | 1549 (11.8%) | 1039 (7.9%) | 887 (6.8%) | 774 (5.9%) |
| openmed-nemotron | 1675 (12.8%) | 1344 (10.3%) | 1334 (10.2%) | 1333 (10.2%) |
| openai-base | 1698 (13.0%) | 1662 (12.7%) | 1662 (12.7%) | 1662 (12.7%) |
| mmbert32k | 1703 (13.0%) | 172 (1.3%) | 44 (0.3%) | 35 (0.3%) |
| openmed-multilingual | 2016 (15.4%) | 1724 (13.2%) | 1719 (13.1%) | 1719 (13.1%) |
| bardsai-eu | 2378 (18.2%) | 1670 (12.8%) | 1516 (11.6%) | 1475 (11.3%) |
| gliner2-hivetrace-omni-ru | 2402 (18.3%) | 1552 (11.9%) | 1282 (9.8%) | 993 (7.6%) |
| gliner2-hivetrace-omni | 2632 (20.1%) | 1688 (12.9%) | 1395 (10.7%) | 1105 (8.4%) |
| nuner-zero | 2767 (21.1%) | 912 (7.0%) | 283 (2.2%) | 37 (0.3%) |
| gliner-pii-edge | 2951 (22.5%) | 458 (3.5%) | 118 (0.9%) | 4 (0.0%) |
| traciora | 2970 (22.7%) | 2852 (21.8%) | 2839 (21.7%) | 2839 (21.7%) |
| pii-shield-onnx | 3002 (22.9%) | 2222 (17.0%) | 2189 (16.7%) | 2189 (16.7%) |
| gliner2-large | 3202 (24.5%) | 1504 (11.5%) | 1081 (8.3%) | 840 (6.4%) |
| kalyan-ettin | 3207 (24.5%) | 2858 (21.8%) | 2843 (21.7%) | 2843 (21.7%) |
| opf-ru | 3226 (24.6%) | 2956 (22.6%) | 2925 (22.3%) | 2922 (22.3%) |
| gliner2-vladlinv-ru | 3449 (26.3%) | 3445 (26.3%) | 3443 (26.3%) | 3440 (26.3%) |
| opf-ru-v2 | 3468 (26.5%) | 3387 (25.9%) | 3380 (25.8%) | 3378 (25.8%) |
| gliner2-vladlinv | 3476 (26.5%) | 3448 (26.3%) | 3439 (26.3%) | 3434 (26.2%) |
| gliner25-fastino-ru | 3988 (30.5%) | 3286 (25.1%) | 3037 (23.2%) | 2664 (20.3%) |
| gliner25-fastino | 4548 (34.7%) | 4105 (31.4%) | 3868 (29.5%) | 3542 (27.1%) |
| gliner-urchade | 4657 (35.6%) | 3150 (24.1%) | 2578 (19.7%) | 1847 (14.1%) |
| gliner-nvidia | 5022 (38.4%) | 3911 (29.9%) | 3368 (25.7%) | 2705 (20.7%) |
| gliner-stream-pii | 5617 (42.9%) | 3944 (30.1%) | 3065 (23.4%) | 2088 (15.9%) |
| gliner-urchade-ru | 5664 (43.3%) | 4296 (32.8%) | 3452 (26.4%) | 2333 (17.8%) |
| gliner-nvidia-ru | 6820 (52.1%) | 5570 (42.5%) | 4747 (36.3%) | 3813 (29.1%) |
| gravitee-small | 7163 (54.7%) | 6861 (52.4%) | 6861 (52.4%) | 6861 (52.4%) |
| gliner2-hivetrace-uni | 7628 (58.3%) | 5187 (39.6%) | 3719 (28.4%) | 2428 (18.5%) |
| fef2-secret-ru | 7959 (60.8%) | 7923 (60.5%) | 7923 (60.5%) | 7923 (60.5%) |
| ner-ru-gherman | 7972 (60.9%) | 7943 (60.7%) | 7942 (60.7%) | 7942 (60.7%) |
| gliner-pii-base | 8777 (67.0%) | 2886 (22.0%) | 1033 (7.9%) | 695 (5.3%) |
| gliner-multi-v21 | 10169 (77.7%) | 6517 (49.8%) | 4547 (34.7%) | 2446 (18.7%) |
| davlan-xlmr | 10468 (79.9%) | 10430 (79.7%) | 10430 (79.7%) | 10430 (79.7%) |
| ner-ru-yqelz | 10705 (81.8%) | 10620 (81.1%) | 10620 (81.1%) | 10620 (81.1%) |
| davlan-mbert | 11209 (85.6%) | 11209 (85.6%) | 11209 (85.6%) | 11209 (85.6%) |
| gliner-multi-v21-ru | 11271 (86.1%) | 7310 (55.8%) | 5004 (38.2%) | 2316 (17.7%) |
| gliner2-hivetrace-uni-ru | 12306 (94.0%) | 8485 (64.8%) | 5569 (42.5%) | 3031 (23.1%) |

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.