Image Quality Understanding on Q-Bench subset (dev)
85.82Yes/No AccuracyQualiRAG
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| QualiRAGRequires VQA-labeled data=false2026.01 | 85.82 | 81.74 | 84.6 | 74.59 | 77.01 | 81.71 | 85.2 | 87.35 | |
| VisualQuality-R1Requires VQA-labeled data=true2026.01 | 82.91 | 79 | 82.3 | 71.6 | 77.63 | 77.78 | 79.61 | 83.27 | |
| Gemini-3-ProRequires VQA-labeled data=false2026.01 | 82.91 | 80.63 | 88.48 | 71.05 | 77.33 | 78.83 | 83.8 | 86.67 | |
| GPT-5Requires VQA-labeled data=false2026.01 | 82.18 | 80.27 | 84.51 | 74.24 | 77.63 | 79.86 | 80.92 | 85.71 | |
| Co-InstructRequires VQA-labeled data=true2026.01 | 82 | 75.38 | 76.77 | 66.73 | 75.29 | 71.53 | 74.67 | 83.27 | |
| Q-InsightRequires VQA-labeled data=true2026.01 | 82 | 77.66 | 82.3 | 68.56 | 76.07 | 76.39 | 77.3 | 83.67 | |
| Qwen2.5-VL-7BRequires VQA-labeled data=false2026.01 | 81.45 | 77.86 | 83.19 | 68.97 | 75.1 | 76.85 | 77.96 | 85.31 | |
| GLM-4.1V-9BRequires VQA-labeled data=false2026.01 | 81.45 | 77.39 | 80.53 | 69.98 | 76.07 | 75.23 | 76.64 | 84.9 | |
| LLaVA-OneVision-7BRequires VQA-labeled data=false2026.01 | 81.27 | 76.79 | 80.31 | 68.56 | 73.74 | 77.78 | 74.67 | 84.08 | |
| Qwen3-VL-8BRequires VQA-labeled data=false2026.01 | 80.91 | 78.53 | 83.85 | 70.99 | 77.43 | 77.08 | 78.62 | 83.27 | |
| InternVL3.5-8BRequires VQA-labeled data=false2026.01 | 80.36 | 76.72 | 79.87 | 69.78 | 75.1 | 74.54 | 76.64 | 84.08 | |
| InternVL3-9BRequires VQA-labeled data=false2026.01 | 80.18 | 75.99 | 78.76 | 68.76 | 72.76 | 75.69 | 73.68 | 86.12 | |
| Q-InstructRequires VQA-labeled data=true2026.01 | 76.91 | 66.35 | 65.04 | 55.78 | 64.01 | 67.13 | 64.8 | 71.84 | |
| mPLUG-Owl3-7BRequires VQA-labeled data=false2026.01 | 63.48 | 63.4 | 77.85 | 50.87 | 57.8 | 66.89 | 67.33 | 64 |