ai4data/gliner2_datause

1---2license: apache-2.03pipeline_tag: token-classification4tags:5  - ner6  - gliner27  - data-use8---9 10# gliner2_datause_extended11 12Fine-tune of `fastino/gliner2-large-v1` (GLiNER2) for data-use mention extraction13(dataset / survey / census / registry mentions in economics research papers).14 15## Labels16- `NAMED_DATA` — a proper name, title, or acronym of a specific data source17- `DESCRIPTIVE_DATA` — a source described in words but not named18- `VAGUE_DATA` — generic data wording with no identifiable source19 20## Training21- base model: `fastino/gliner2-large-v1`22- dataset: `rafmacalaba/data-use-mentions-extended` (gliner2 config)23- epochs: 524- encoder LR: 1e-0525- task LR: 0.000526- batch size: 827- precision: bf1628 29## Evaluation (holdout, label-agnostic)30 31| thr | tp | fp | fn | precision | recall | f0.5 | f1 |32| --- | --- | --- | --- | --- | --- | --- | --- |33| 0.10 | 12163 | 5439 | 404 | 0.6910 | 0.9679 | 0.7329 | 0.8063 |34| 0.20 | 11976 | 4277 | 591 | 0.7368 | 0.9530 | 0.7719 | 0.8311 |35| 0.30 | 11788 | 3535 | 779 | 0.7693 | 0.9380 | 0.7980 | 0.8453 |36| 0.40 | 11618 | 2942 | 949 | 0.7979 | 0.9245 | 0.8204 | 0.8566 |37| 0.50 | 11383 | 2478 | 1184 | 0.8212 | 0.9058 | 0.8368 | 0.8614 |38| 0.60 | 11068 | 2008 | 1499 | 0.8464 | 0.8807 | 0.8531 | 0.8632 |39| 0.70 | 10487 | 1535 | 2080 | 0.8723 | 0.8345 | 0.8645 | 0.8530 |40 41**Best F0.5**: 0.8645 (thr=0.7)42**Best F1**: 0.8632 (thr=0.6)43 44<!-- NER_COMPARISON_START -->45## NER holdout comparison46device: NVIDIA H100 NVL47 48### rafmacalaba/data-use-mentions-extended (n=9249)49 50| model | backend | best F0.5 | thr | best F1 | thr | wall-clock (s) | texts/s |51| --- | --- | --- | --- | --- | --- | --- | --- |52| `rafmacalaba/gliner_datause_extended` | gliner | 0.8506 | 0.6 | 0.8549 | 0.5 | 202.7 | 45.6 |53| `ai4data/gliner2_datause` | gliner2 | 0.8634 | 0.7 | 0.8624 | 0.6 | 180.8 | 51.1 |54 55F0.5 by threshold (sweet spots side-by-side):56 57| thr | `rafmacalaba/gliner_datause_extended` | `ai4data/gliner2_datause` |58| --- | --- | --- |59| 0.1 | 0.6476 | 0.7321 |60| 0.2 | 0.7121 | 0.7712 |61| 0.3 | 0.7505 | 0.7979 |62| 0.4 | 0.7858 | 0.8201 |63| 0.5 | 0.8224 | 0.8363 |64| 0.6 | 0.8506 | 0.8523 |65| 0.7 | 0.8422 | 0.8634 |66<!-- NER_COMPARISON_END -->67 68