Multi-task Language Understanding on MMLU-Pro (Δ%)
5.31MMLU-Pro Delta (%)MedGemma
Evaluation Results
| Method | Links | |
|---|---|---|
| MedGemmaSize=4B, Variant=Med2026.04 | 5.31 | |
| MedGemmaSize=27B, Variant=Med2026.04 | 4.91 | |
| Qwen-2.5VLSize=32B, Variant=FT2026.04 | 2.25 | |
| Gemma-3Size=4B, Variant=FT2026.04 | 1.87 | |
| Gemma-3Size=27B, Variant=FT2026.04 | 1.04 | |
| Gemma-3Size=12B, Variant=FT2026.04 | 0.54 | |
| Qwen-2.5VLSize=7B, Variant=FT2026.04 | 0.02 |