BioMedicine on BioMedicine Domain Bench
58PubMedQA ScoreBiomedicine
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| BiomedicineMerging Method=Expert Model (MMed-Llama-8B)2026.01 | 58 | 56.41 | 34.03 | |
| ReasonAnyMerging Method=ReasonAny (ours)2026.01 | 56.4 | 48.88 | 56.96 | |
| DAREMerging Method=DARE Merging2026.01 | 54.4 | 8.27 | 35.27 | |
| ReasoningMerging Method=Expert Model (DeepSeek-R1-Distill-Llama-8B)2026.01 | 51.5 | 35.4 | 57.45 | |
| TIESMerging Method=TIES Merging2026.01 | 48 | 11.71 | 35.08 | |
| Task ArithmeticMerging Method=Task Arithmetic2026.01 | 46.4 | 30.3 | 42.71 | |
| LinearMerging Method=Linear Merging2026.01 | 36.8 | 27.51 | 35.07 | |
| LEDMerging Method=LED Merging2026.01 | 15.6 | 48.51 | 41.34 | |
| FuseLLMMerging Method=FuseLLM2026.01 | 11 | 6.78 | 22.15 |