Medical Visual Question Answering on PMC-VQA OOD (test)
0.058ECEOurs
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| OursBackbone=Qwen2-VL-7B-Instruct, Trained on=OmniMedVQA2026.06 | 0.058 | 0.216 | 70.3 | 49 | |
| OursModel Architecture=MedGemma-4B-IT, Training Pathway=OmniMedVQA-trained2026.06 | 0.097 | 0.243 | 0.64 | 46.3 | |
| SteerConfBackbone=Qwen2-VL-7B-Instruct, Trained on=OmniMedVQA2026.06 | 0.199 | 0.264 | 69.2 | 48.4 | |
| ConfTunerBackbone=Qwen2-VL-7B-Instruct, Trained on=OmniMedVQA2026.06 | 0.229 | 0.295 | 67.2 | 50.3 | |
| ConfTunerModel Architecture=MedGemma-4B-IT, Training Pathway=OmniMedVQA-trained2026.06 | 0.299 | 0.329 | 0.583 | 43.6 | |
| SteerConfModel Architecture=MedGemma-4B-IT, Training Pathway=OmniMedVQA-trained2026.06 | 0.302 | 0.336 | 0.615 | 46.2 | |
| Top-K SamplingBackbone=Qwen2-VL-7B-Instruct, Trained on=OmniMedVQA2026.06 | 0.377 | 0.381 | 61.7 | 44 | |
| Base ModelModel Architecture=MedGemma-4B-IT, Training Pathway=OmniMedVQA-trained2026.06 | 0.445 | 0.445 | 0.533 | 45.5 | |
| Base ModelBackbone=Qwen2-VL-7B-Instruct, Trained on=OmniMedVQA2026.06 | 0.504 | 0.503 | 51.7 | 48.8 | |
| Top-K SamplingModel Architecture=MedGemma-4B-IT, Training Pathway=OmniMedVQA-trained2026.06 | 0.569 | 0.541 | 0.535 | 31.9 |