Visual Question Answering on VQA-RAD (test)
90.4Overall AccuracyVILA-M3
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| VILA-M3Number of Parameters=40B2024.11 | 90.4 | — | — | — | |
| VILA-M3Number of Parameters=8B2024.11 | 84.7 | — | — | — | |
| Task Spfc. SOTA2024.11 | 84.2 | — | — | — | |
| LCT2026.03 | 81.2 | 70.9 | 87.9 | — | |
| VILA-M3Number of Parameters=13B2024.11 | 80.5 | — | — | — | |
| CIMB-MVQA2026.03 | 79.4 | 69.3 | 86.2 | — | |
| Med-GeminiNumber of Parameters=1.5T2024.11 | 78.8 | — | — | — | |
| DeCoCT2026.03 | 78.3 | 67.1 | 85.7 | — | |
| VILA-M3Number of Parameters=3B2024.11 | 78.2 | — | — | — | |
| PMC-CLIP2026.03 | 77.6 | 67 | 84 | — | |
| M3AEvision encoder=CLIP-ViT-B, language encoder=RoBERTa-base2022.09 | 77.01 | 67.23 | 83.46 | — | |
| M3AE2026.03 | 77 | 67.2 | 83.5 | — | |
| MISS2026.03 | 76.1 | 71.8 | 80.4 | — | |
| CCIS-MVQA2026.03 | 75.1 | 68.8 | 79.2 | — | |
| M2I22026.03 | 73.7 | 61.8 | 81.6 | — | |
| DeBCF2026.03 | 71.6 | 58.6 | 80.9 | — | |
| CLIPQCR2026.03 | 71.1 | 58 | 79.6 | — | |
| CPRD-BAN2022.09 | 67.8 | 52.5 | 77.9 | — | |
| CPRD+BAN2026.03 | 67.8 | 52.5 | 77.9 | — | |
| MMQAttention Method=BAN, refinement steps (m)=5, meta-models (n)=32021.05 | 67 | 53.7 | 75.8 | — | |
| MEVF-BAN2022.09 | 66.1 | 49.2 | 77.2 | — | |
| MEVF+BAN2026.03 | 66.1 | 49.2 | 77.2 | — | |
| MEVF-SAN2022.09 | 64.1 | 49.2 | 73.9 | — | |
| MEVF+SAN2026.03 | 64.1 | 49.2 | 73.9 | — | |
| MMQAttention Method=SAN, refinement steps (m)=5, meta-models (n)=32021.05 | 64 | 46.3 | 75.7 | — | |
| MEVFAttention Method=BAN2021.05 | 62.7 | 43.9 | 75.1 | — | |
| MEVFAttention Method=SAN2021.05 | 60.7 | 40.7 | 74.1 | — | |
| Qwen2-VLSize=7B, Source=Open2025.02 | 60.5 | 43 | 74.5 | — | |
| MAMLAttention Method=BAN2021.05 | 59.6 | 40.1 | 72.4 | — | |
| Llama3.2-visionSize=11B, Source=Open2025.02 | 59 | 51.5 | 64.9 | — | |
| BAN2022.09 | 58.3 | 37.4 | 72.1 | — | |
| InternVL2Size=8B, Source=Open2025.02 | 58.3 | 48 | 66.5 | — | |
| IDEFICS2Size=8B, Source=Open2025.02 | 57.4 | 45 | 67.3 | — | |
| Claude-3.5Source=Closed2025.02 | 57.2 | 51 | 62.2 | — | |
| MAMLAttention Method=SAN2021.05 | 57.1 | 38.2 | 69.7 | — | |
| SAN2022.09 | 54.3 | 31.3 | 69.5 | — | |
| GPT-4oSource=Closed2025.02 | 53.4 | 48 | 57.8 | — | |
| InternVL2Size=4B, Source=Open2025.02 | 52.5 | 37 | 64.9 | — | |
| BIANAttention Method=BAN, ImageNet pre-trained=true2021.05 | 52.3 | 28.4 | 67.9 | — | |
| InternVL2Size=2B, Source=Open2025.02 | 51.2 | 35 | 64.1 | — | |
| MFB2022.09 | 50.6 | 14.5 | 74.3 | — | |
| Qwen2-VLSize=2B, Source=Open2025.02 | 49.9 | 38 | 59.4 | — | |
| MMGUARD-BPH-MaxTarget Model=Gemma2026.05 | 47.5 | — | — | 4.5 | |
| DeepSeek-VLSize=7B, Source=Open2025.02 | 45.7 | 30 | 58.2 | — | |
| StANAttention Method=SAN, ImageNet pre-trained=true2021.05 | 44.2 | 24.2 | 57.2 | — | |
| DeepSeek-VLSize=1.3B, Source=Open2025.02 | 42.6 | 25 | 56.6 | — | |
| InternVL2Size=1B, Source=Open2025.02 | 39 | 26.5 | 49 | — | |
| Zero-Shot (Clean)Target Model=Gemma2026.05 | 7.3 | — | — | 44.7 | |
| CPRD-BAN2025.01 | — | 52.5 | 77.9 | — | |
| M3AE2025.01 | — | 67.2 | 83.5 | — | |
| MedVInT2025.01 | — | 69.3 | 84.2 | — | |
| MEVF-BAN2025.01 | — | 49.2 | 77.2 | — | |
| PMC-CLIP2025.01 | — | 67 | 84 | — | |
| UMed-LVLM2025.01 | — | 74.9 | 87.6 | — |