datause-relation-v0

Fine-tuned GLiNER2 adapter for relation extraction in Data Use impact assessment pipeline.

Architecture: LoRA adapter for fastino/gliner2-large-v1
Relations: has_organization, used_by, has_acronym, has_timeframe, has_geography
Training data: LLM-generated synthetic + real-world PDF text
Status: v0 alpha — initial fine-tuning.

Usage

from gliner import GLiNER2

model = GLiNER2.from_pretrained("fastino/gliner2-large-v1")
model.load_adapter("ai4data/datause-relation-v0")

Pipeline Integration

Call 1b in 3-model swarm:

• Call 1: Entity extraction (ai4data/datause-extraction)
• Call 1b: Relation extraction (this model)
• Call 2: Classification (ai4data/datause-impact-v0)