gliner_datause_smoketest

Fine-tune of urchade/gliner_large-v2.1 for data-use mention extraction (dataset / survey / census / registry mentions in economics research papers).

Labels

• NAMED_DATA — a proper name, title, or acronym of a specific data source
• DESCRIPTIVE_DATA — a source described in words but not named
• VAGUE_DATA — generic data wording with no identifiable source

Training

• base model: urchade/gliner_large-v2.1
• dataset: rafmacalaba/data-use-mentions (gliner config)
• epochs: 1
• learning rate: 5e-06
• batch size: 16
• precision: bf16

Evaluation (holdout)

thr	tp	fp	fn	precision	recall	f0.5	f1
0.10	71	1253	20	0.0536	0.7802	0.0659	0.1004
0.20	54	519	37	0.0942	0.5934	0.1133	0.1627
0.30	29	204	62	0.1245	0.3187	0.1417	0.1790
0.40	21	67	70	0.2386	0.2308	0.2370	0.2346
0.50	11	34	80	0.2444	0.1209	0.2030	0.1618
0.60	4	23	87	0.1481	0.0440	0.1005	0.0678
0.70	1	15	90	0.0625	0.0110	0.0323	0.0187

Best F0.5: 0.2370 (thr=0.4) Best F1: 0.2346 (thr=0.4)