Visual Question Answering on VQA-RAD open-ended
29Exact Match (EM)Greedy
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| GreedyModel=MedGemma 4B2026.05 | 29 | 38.38 | 65.3 | 54 | |
| BDG-WModel=MedGemma 4B2026.05 | 28.9 | 40.25 | 65.71 | 56.7 | |
| SCDModel=MedGemma 4B2026.05 | 28.8 | 38.77 | 65.72 | 55.2 | |
| BDG-ClassicModel=MedGemma 4B2026.05 | 28.4 | 39.69 | 65.54 | 56.4 | |
| VerifierModel=MedGemma 4B2026.05 | 26.3 | 37.48 | 64.47 | 54.8 | |
| GreedyModel=Qwen3-VL 8B2026.05 | 21 | 31.03 | 62.38 | 40.5 | |
| BDG-WModel=Qwen3-VL 8B2026.05 | 20.5 | 30.88 | 62.59 | 39.29 | |
| BDG-ClassicModel=Qwen3-VL 8B2026.05 | 20.2 | 30.6 | 62.52 | 39.47 | |
| BDG-WModel=Qwen3-VL 4B2026.05 | 20.1 | 29.47 | 61.93 | 40.3 | |
| BDG-WModel=Qwen3-VL 2B2026.05 | 19.66 | 28.83 | 60.61 | 37.04 | |
| BDG-ClassicModel=Qwen3-VL 4B2026.05 | 19.1 | 28.37 | 61.54 | 39.4 | |
| SCDModel=Qwen3-VL 8B2026.05 | 19.1 | 29.78 | 61.52 | 37.09 | |
| BDG-ClassicModel=Qwen3-VL 2B2026.05 | 19.06 | 28.58 | 60.6 | 36.51 | |
| VerifierModel=Qwen3-VL 8B2026.05 | 18.89 | 30 | 62.18 | 39.8 | |
| VerifierModel=Qwen3-VL 4B2026.05 | 18 | 28.04 | 61.44 | 40.9 | |
| GreedyModel=Qwen3-VL 2B2026.05 | 17.5 | 26.73 | 59.36 | 33.5 | |
| VerifierModel=Qwen3-VL 2B2026.05 | 17.25 | 27.71 | 60.82 | 35.71 | |
| SCDModel=Qwen3-VL 4B2026.05 | 16.4 | 24.98 | 59.92 | 36.6 | |
| GreedyModel=Qwen3-VL 4B2026.05 | 15.5 | 25.63 | 59.74 | 36.5 | |
| SCDModel=Qwen3-VL 2B2026.05 | 15.35 | 23.33 | 58.62 | 30.79 | |
| BDG-ClassicModel=Gemma-3 4B2026.05 | 13 | 23.58 | 57.48 | 42.7 | |
| BDG-WModel=Gemma-3 4B2026.05 | 12.5 | 23.08 | 57.2 | 42.3 | |
| GreedyModel=Gemma-3 4B2026.05 | 12 | 24.03 | 56.91 | 40 | |
| SCDModel=Gemma-3 4B2026.05 | 9.5 | 20.81 | 55.35 | 37.9 | |
| VerifierModel=Gemma-3 4B2026.05 | 8.2 | 20.16 | 55.67 | 40.8 |