Computer Graphics Quality Assessment on R4-CGQA (test)
79.21Choice AccuracyQwen 2.5-VL-32B
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Qwen 2.5-VL-32BConfiguration=R4-CGQA2026.03 | 79.21 | 70.24 | 2.87 | 1.95 | |
| Qwen 2.5-VL-32BConfiguration=Original2026.03 | 77.71 | 67.5 | 2.79 | — | |
| LlavaNext-32BConfiguration=R4-CGQA2026.03 | 75.84 | 68.72 | 2.9 | 2.58 | |
| LlavaNext-32BConfiguration=Original2026.03 | 75 | 63.21 | 2.83 | — | |
| Pixtral-12BConfiguration=R4-CGQA2026.03 | 73.12 | 69.1 | 2.89 | 2.7 | |
| LlavaNext-8BConfiguration=R4-CGQA2026.03 | 72.74 | 67.43 | 2.72 | 3.21 | |
| Pixtral-12BConfiguration=Original2026.03 | 70.11 | 66.2 | 2.78 | — | |
| LlavaNext-8BConfiguration=Original2026.03 | 69.41 | 61.92 | 2.68 | — | |
| Qwen 2.5-VL-7BConfiguration=R4-CGQA2026.03 | 67.76 | 63.87 | 2.46 | 8.96 | |
| MiniCPM-V-8BConfiguration=R4-CGQA2026.03 | 67.63 | 61.98 | 2.34 | 8.16 | |
| Llama 3.2-V-11BConfiguration=R4-CGQA2026.03 | 67.28 | 67.26 | 2.31 | 6.89 | |
| Gemma3-4BConfiguration=R4-CGQA2026.03 | 66.82 | 65.22 | 2.32 | 12.62 | |
| Qwen 2.5-VL-7BConfiguration=Original2026.03 | 66.73 | 61.21 | 1.3 | — | |
| Gemma3-4BConfiguration=Original2026.03 | 66.03 | 53.55 | 1.05 | — | |
| Llama 3.2-V-11BConfiguration=Original2026.03 | 64.59 | 56.87 | 1.93 | — | |
| LLava-1.6-13BConfiguration=R4-CGQA2026.03 | 61.43 | 60.37 | 2.56 | 7.77 | |
| MiniCPM-V-8BConfiguration=Original2026.03 | 60.05 | 53.47 | 1.92 | — | |
| LLava-1.6-7BConfiguration=R4-CGQA2026.03 | 58.06 | 58.84 | 2.5 | 7.16 | |
| Bakllava-7BConfiguration=R4-CGQA2026.03 | 55.97 | 61.17 | 1.96 | 8.79 | |
| LLava-1.6-13BConfiguration=Original2026.03 | 53.96 | 51.34 | 2.22 | — | |
| LLava-1.6-7BConfiguration=Original2026.03 | 51.7 | 49.73 | 2.2 | — | |
| Bakllava-7BConfiguration=Original2026.03 | 43.72 | 52.85 | 1.67 | — |