gliner_datause_v0

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-v2 (gliner config)
• corpus: all
• epochs: 5
• learning rate: 5e-06
• batch size: 16
• precision: bf16

Evaluation (holdout)

thr	tp	fp	fn	precision	recall	f0.5	f1
0.10	16553	64389	552	0.2045	0.9677	0.2428	0.3377
0.20	16274	32132	831	0.3362	0.9514	0.3861	0.4968
0.30	15597	18446	1508	0.4582	0.9118	0.5088	0.6099
0.40	13992	10141	3113	0.5798	0.8180	0.6156	0.6786
0.50	10189	4475	6916	0.6948	0.5957	0.6724	0.6414
0.60	4852	1332	12253	0.7846	0.2837	0.5798	0.4167
0.70	1272	245	15833	0.8385	0.0744	0.2745	0.1366

Best F0.5: 0.6724 (thr=0.5) Best F1: 0.6786 (thr=0.4)

Evaluation breakdown (holdout)

group	examples	spans	thr	precision	recall	f0.5	f1
overall	15332	17332	0.50	0.6948	0.5957	0.6724	0.6414
prwp	7758	11375	0.50	0.7719	0.6282	0.7382	0.6927
fcv	7574	5957	0.50	0.5642	0.5318	0.5574	0.5475
general_prwp	7758	11375	0.50	0.7719	0.6282	0.7382	0.6927
fcv_pads_east_africa	5351	3920	0.50	0.5332	0.5332	0.5332	0.5332
jdc_operational	145	110	0.50	0.4776	0.5872	0.4961	0.5267
refugee_pads	573	377	0.50	0.4732	0.5442	0.4859	0.5062
reliefweb	1505	1550	0.50	0.7178	0.5212	0.6675	0.6039

Per-label (overall)

label	examples	spans	thr	precision	recall	f0.5	f1
NAMED_DATA	15332	7107	0.60	0.6049	0.4388	0.5623	0.5087
DESCRIPTIVE_DATA	15332	8601	0.50	0.4997	0.2891	0.4361	0.3663
VAGUE_DATA	15332	1624	0.40	0.2019	0.1078	0.1719	0.1405

Origin breakdown (per-origin metrics)

origin	examples	spans	thr	precision	recall	f0.5	f1
fcv_pads_east_africa	5351	3920	0.50	0.5332	0.5332	0.5332	0.5332
general_prwp	7758	11375	0.50	0.7719	0.6282	0.7382	0.6927
jdc_operational	145	110	0.50	0.4776	0.5872	0.4961	0.5267
refugee_pads	573	377	0.50	0.4732	0.5442	0.4859	0.5062
reliefweb	1505	1550	0.50	0.7178	0.5212	0.6675	0.6039

Per-label details

label	examples	spans	thr	precision	recall	f0.5	f1
NAMED_DATA	15332	7107	0.60	0.6049	0.4388	0.5623	0.5087
DESCRIPTIVE_DATA	15332	8601	0.50	0.4997	0.2891	0.4361	0.3663
VAGUE_DATA	15332	1624	0.40	0.2019	0.1078	0.1719	0.1405