Visualization Assessment on VISJUDGE-BENCH
0.421Overall MAEVISJUDGE
Evaluation Results
| Method | Links | |||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| VISJUDGEType=VisJudge (Ours), Backbone=Qwen2.5-VL-7B-Instruct2025.10 | 0.421 | 0.661 | 0.648 | 0.677 | 0.58 | 0.545 | 0.604 | 0.286 | 0.747 | 0.69 | 0.756 | 0.545 | 0.498 | 0.578 | 0.687 | 0.574 | 0.628 | 0.576 | 0.568 | 0.513 | 0.385 | |
| VISJUDGEType=VisJudge (Ours), Backbone=Qwen2.5-VL-3B-Instruct2025.10 | 0.491 | 0.705 | 0.721 | 0.696 | 0.625 | 0.571 | 0.616 | 0.377 | 0.855 | 0.85 | 0.822 | 0.62 | 0.539 | 0.603 | 0.648 | 0.533 | 0.581 | 0.579 | 0.504 | 0.49 | 0.402 | |
| VISJUDGEType=VisJudge (Ours), Backbone=Llava-v1.6-mistral-7B2025.10 | 0.496 | 0.767 | 0.695 | 0.775 | 0.643 | 0.574 | 0.598 | 0.383 | 0.919 | 0.752 | 0.952 | 0.642 | 0.541 | 0.619 | 0.605 | 0.226 | 0.536 | 0.443 | 0.432 | 0.406 | 0.403 | |
| VISJUDGEType=VisJudge (Ours), Backbone=InternVL2-8B2025.10 | 0.541 | 0.797 | 0.769 | 0.691 | 0.608 | 0.645 | 0.616 | 0.437 | 1.068 | 0.951 | 0.822 | 0.638 | 0.736 | 0.655 | 0.66 | 0.533 | 0.594 | 0.545 | 0.499 | 0.391 | 0.42 | |
| GPT-4-TurboType=Closed-source2025.10 | 0.553 | 0.862 | 0.781 | 0.778 | 0.649 | 0.699 | 0.682 | 0.486 | 1.219 | 0.989 | 0.97 | 0.72 | 0.862 | 0.81 | 0.428 | 0.255 | 0.439 | 0.382 | 0.463 | 0.276 | 0.295 | |
| GPT-4oType=Closed-source2025.10 | 0.61 | 0.988 | 0.806 | 0.744 | 0.609 | 0.695 | 0.657 | 0.577 | 1.562 | 1.063 | 0.921 | 0.627 | 0.823 | 0.729 | 0.482 | 0.381 | 0.539 | 0.442 | 0.471 | 0.277 | 0.363 | |
| Claude-3-SonnetType=Closed-source2025.10 | 0.622 | 0.841 | 0.757 | 0.832 | 0.679 | 0.734 | 0.785 | 0.603 | 1.182 | 0.973 | 1.147 | 0.774 | 0.934 | 1.037 | 0.465 | 0.393 | 0.55 | 0.452 | 0.421 | 0.163 | 0.228 | |
| Gemini-1.5-ProType=Closed-source2025.10 | 0.662 | 1.243 | 0.945 | 0.9 | 0.84 | 0.918 | 0.98 | 0.677 | 2.291 | 1.478 | 1.371 | 1.112 | 1.325 | 1.46 | 0.265 | 0.178 | 0.379 | 0.353 | 0.445 | 0.193 | 0.208 | |
| Gemini-1.5-FlashType=Closed-source2025.10 | 0.682 | 0.829 | 0.912 | 0.82 | 0.638 | 0.729 | 0.798 | 0.72 | 1.182 | 1.326 | 1.12 | 0.673 | 0.925 | 1.057 | 0.395 | 0.372 | 0.459 | 0.417 | 0.459 | 0.157 | 0.209 | |
| Qwen2.5-VL-72B-InstructType=Open-source2025.10 | 0.702 | 0.999 | 0.926 | 0.811 | 0.663 | 0.735 | 0.717 | 0.762 | 1.626 | 1.373 | 1.092 | 0.713 | 0.932 | 0.87 | 0.44 | 0.331 | 0.479 | 0.416 | 0.435 | 0.165 | 0.251 | |
| Qwen2.5-VL-32B-InstructType=Open-source2025.10 | 0.703 | 0.909 | 0.987 | 0.801 | 0.678 | 0.761 | 0.719 | 0.756 | 1.361 | 1.528 | 1.067 | 0.728 | 0.988 | 0.88 | 0.435 | 0.348 | 0.468 | 0.408 | 0.449 | 0.2 | 0.268 | |
| Llava-v1.6-mistral-7BType=Open-source2025.10 | 0.724 | 0.929 | 1.124 | 0.939 | 0.801 | 0.814 | 0.818 | 0.817 | 1.513 | 1.96 | 1.403 | 1.007 | 1.071 | 1.076 | 0.18 | 0.201 | 0.137 | 0.16 | 0.188 | 0.064 | 0.076 | |
| InternVL2-8BType=Open-source2025.10 | 0.793 | 1.234 | 1.193 | 0.87 | 0.679 | 0.759 | 0.753 | 0.934 | 2.24 | 2.088 | 1.219 | 0.747 | 0.981 | 0.968 | 0.409 | 0.323 | 0.407 | 0.344 | 0.419 | 0.216 | 0.17 | |
| Qwen2.5-VL-3B-InstructType=Open-source2025.10 | 0.821 | 1.085 | 1.258 | 1.088 | 0.723 | 0.727 | 0.808 | 1.028 | 1.872 | 2.4 | 1.852 | 0.868 | 0.903 | 1.046 | 0.272 | 0.199 | 0.222 | 0.275 | 0.338 | 0.13 | 0.155 | |
| Claude-3.5-SonnetType=Closed-source2025.10 | 0.824 | 0.978 | 0.904 | 1.154 | 0.783 | 0.94 | 0.862 | 1.009 | 1.577 | 1.306 | 1.989 | 0.994 | 1.467 | 1.198 | 0.395 | 0.325 | 0.492 | 0.365 | 0.455 | 0.137 | 0.259 | |
| Qwen2.5-VL-7B-InstructType=Open-source2025.10 | 0.847 | 1.171 | 1.296 | 0.858 | 0.756 | 0.812 | 0.772 | 1.045 | 2.051 | 2.413 | 1.177 | 0.938 | 1.093 | 0.996 | 0.341 | 0.341 | 0.352 | 0.281 | 0.357 | 0.149 | 0.155 |