Multimodal Visual Pattern Understanding on MMVP
80.33AccuracyGRPO-CARE
Evaluation Results
| Method | Links | |
|---|---|---|
| GRPO-CARE2025.12 | 80.33 | |
| Rotation + CL + CARE2025.12 | 79.67 | |
| Jigsaw + CL + CARE2025.12 | 78.67 | |
| Mix + CL + CARE2025.12 | 78.33 | |
| Vision-Zero2025.12 | 78 | |
| PatchFit + CL + CARE2025.12 | 78 | |
| Qwen 2.5 VL2025.12 | 77.67 | |
| VisualSphinx2025.12 | 77.33 | |
| Jigsaw + CARE2025.12 | 77 | |
| Visual Jigsaw2025.12 | 76.33 | |
| Jigsaw2025.12 | 76.33 | |
| ViCrit2025.12 | 75.33 | |
| Jigsaw + CL2025.12 | 74.67 | |
| FedAvg-TASetting=MCross2026.03 | 35.6 | |
| FedAvgSetting=MCross2026.03 | 34.8 | |
| Fed-CMPSetting=MCross2026.03 | 34.7 | |
| MOONSetting=MCross2026.03 | 33.8 | |
| FedAdamSetting=MCross2026.03 | 33.1 | |
| FedAvg-TIEsSetting=MCross2026.03 | 32.2 | |
| FedProxSetting=MCross2026.03 | 31.4 | |
| LLaVA 1.5 + I-MoFres=336x336, #tokens=11522024.01 | 31.3 | |
| Local TrainingSetting=MCross2026.03 | 31.2 | |
| LLaVA 1.5 + I-MoFres=224x224, #tokens=5122024.01 | 28 | |
| FedAvg-DARESetting=MCross2026.03 | 27.2 | |
| LLaVA 1.5res=336x336, #tokens=5762024.01 | 24.7 |