Multimodal Evaluation on MMT-Bench
62.65AccuracyMix + CL + CARE
Evaluation Results
| Method | Links | |
|---|---|---|
| Mix + CL + CAREModel Size=7B2025.12 | 62.65 | |
| GRPO-CAREModel Size=7B2025.12 | 62.62 | |
| Jigsaw + CLModel Size=7B2025.12 | 62.62 | |
| Jigsaw + CL + CAREModel Size=7B2025.12 | 62.52 | |
| Rotation + CL + CAREModel Size=7B2025.12 | 61.91 | |
| Jigsaw + CAREModel Size=7B2025.12 | 61.18 | |
| VisualSphinxModel Size=7B2025.12 | 60.63 | |
| Vision-ZeroModel Size=7B2025.12 | 60.47 | |
| JigsawModel Size=7B2025.12 | 59.96 | |
| ViCritModel Size=7B2025.12 | 59.83 | |
| Qwen 2.5 VLModel Size=7B2025.12 | 59.64 | |
| PatchFit + CL + CAREModel Size=7B2025.12 | 58.94 | |
| Visual JigsawModel Size=7B2025.12 | 57.76 |