n-ary drug combination extraction on DrugComb (test)
74.6F1 (pos, exact)RexDrug
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| RexDrugModel=llama3.1-8b-instruct, Category=RexDrug (w/ reasoning)2026.03 | 74.6 | 77.1 | 81.4 | 85.3 | |
| RexDrugModel=qwen2.5-7b-instruct, Category=RexDrug (w/ reasoning)2026.03 | 74.2 | 76.9 | 81.3 | 84.1 | |
| llama3.1-8b-instructCategory=SFT (w/o reasoning)2026.03 | 72.9 | 75.4 | 80.5 | 83.4 | |
| qwen2.5-7b-instructCategory=SFT (w/o reasoning)2026.03 | 72.1 | 74.6 | 80 | 82.9 | |
| RCFIND*Model=PubmedBERT, Category=RCFIND*2026.03 | 72 | 74.9 | 80.3 | 83.3 | |
| qwen2.5-7b-instructCategory=SFT (w/ reasoning)2026.03 | 69.3 | 72.1 | 75.2 | 79.4 | |
| llama3.1-8b-instructCategory=SFT (w/ reasoning)2026.03 | 68.9 | 72.8 | 75.1 | 80.3 | |
| Seq2RelModel=PubmedBERT, Category=Seq2Rel2026.03 | 66.7 | — | 71.1 | — | |
| PURE*Model=PubmedBERT+DAPT, Category=PURE*2026.03 | 61.8 | 67.7 | 69.4 | 77.5 | |
| GPT-4oCategory=Zero-shot (w/o reasoning)2026.03 | 49.2 | 53.6 | 54.5 | 60.6 | |
| qwen3-30b-instructCategory=Zero-shot (w/o reasoning)2026.03 | 46 | 51.9 | 51.6 | 59.1 | |
| GPT-4oCategory=Zero-shot (w/ reasoning)2026.03 | 43.8 | 48 | 50.9 | 55.9 | |
| llama3.1-8b-instructCategory=only-RL (w/ reasoning)2026.03 | 40 | 46.2 | 46.1 | 53.3 | |
| qwen3-30b-thinkCategory=Zero-shot (w/ reasoning)2026.03 | 39.8 | 44.9 | 43 | 47.8 | |
| GPT-3.5Category=Zero-shot (w/o reasoning)2026.03 | 39.4 | 41.9 | 40.7 | 43.4 | |
| qwen2.5-7b-instructCategory=only-RL (w/ reasoning)2026.03 | 38.1 | 46.2 | 43.6 | 52.6 | |
| qwen2.5-7b-instructCategory=Zero-shot (w/o reasoning)2026.03 | 37.1 | 41.2 | 45.3 | 51.5 | |
| qwen3-30b-instructCategory=Zero-shot (w/ reasoning)2026.03 | 36.2 | 43.2 | 40.1 | 49 | |
| llama3.1-8b-instructCategory=Zero-shot (w/o reasoning)2026.03 | 34.7 | 40.1 | 42.4 | 49.5 | |
| llama3.1-8b-instructCategory=Zero-shot (w/ reasoning)2026.03 | 30.2 | 38.3 | 36.4 | 44.9 | |
| qwen2.5-7b-instructCategory=Zero-shot (w/ reasoning)2026.03 | 29.5 | 34.7 | 34.3 | 40.5 | |
| GPT-3.5Category=Zero-shot (w/ reasoning)2026.03 | 27.9 | 31.8 | 29.9 | 31.5 |