Classification on MedNLI 200 random samples (test)
88.6Macro F1 ScoreCAPT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CAPTComponents=Mnew + Mold-clin + Mold2026.01 | 88.6 | 88.5 | |
| UniTEComponents=Mnew + Mold-clin2026.01 | 87.3 | 87.5 | |
| Qwen3-30BModel Category=New-gen general (Mnew)2026.01 | 85.8 | 85 | |
| LLaMA-2-70B-chatModel Category=Old-gen large general (Mold-L)2026.01 | 59.8 | 63.5 | |
| Proxy TuningComponents=Mold-L + Mold-clin + Mold2026.01 | 57.5 | 62 | |
| MeLLaMA-13B-chatModel Category=Old-gen clinical (Mold-clin)2026.01 | 53.9 | 53.5 |