Medical Visual Question Answering on OmniMed-VQA
0.835BERTScoreRform + Rexact + RDTW
Evaluation Results
| Method | Links | |
|---|---|---|
| Rform + Rexact + RDTWTraining Strategy=Formulaic + Exact Match + DTW reward training2026.04 | 0.835 | |
| SFTTraining Strategy=Supervised Fine-Tuning2026.04 | 0.833 | |
| Rform + RexactTraining Strategy=Formulaic + Exact Match reward training2026.04 | 0.824 | |
| RformTraining Strategy=Formulaic reward training2026.04 | 0.784 |