NER-RE on 53 lung pathology reports (held-out)
96Micro F1NER_RE (Baseline)
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| NER_RE (Baseline)Backbone=GatorTron, Learning Paradigm=supervised2026.06 | 96 | 94.4 | 97.7 | 637 | 38 | 15 | |
| GPT-OSS-20Binference=zero-shot, workflow=LangGraph agentic workflow2026.06 | 89.3 | 84.3 | 94.9 | 619 | 115 | 33 | |
| GPT-OSS-120Binference=zero-shot, workflow=LangGraph agentic workflow2026.06 | 88.3 | 82 | 95.7 | 624 | 137 | 28 | |
| LLAMA-3.3-70Binference=zero-shot, workflow=LangGraph agentic workflow2026.06 | 87.2 | 80.4 | 95.4 | 622 | 152 | 30 | |
| GRANITE-3.3-8Binference=zero-shot, workflow=LangGraph agentic workflow2026.06 | 83.4 | 76 | 92.4 | 597 | 189 | 49 | |
| GEMMA-3-27Binference=zero-shot, workflow=LangGraph agentic workflow2026.06 | 83.1 | 74.8 | 93.6 | 610 | 206 | 42 |