Multimodal Understanding on MMMU Pro
55.7Vis AccuracyMiMo-VL-Miloco-7B
Evaluation Results
| Method | Links | |
|---|---|---|
| MiMo-VL-Miloco-7B2025.12 | 55.7 | |
| MiMo-VL-Miloco-7B2025.12 | 47.2 | |
| InternVL3-8B2025.12 | 45.6 | |
| MiMo-VL-7B-SFT-25082025.12 | 43 | |
| Gemma-3-27B-IT2025.12 | 37.8 | |
| InternVL3-8B2025.12 | 37.8 | |
| MiMo-VL-7B-SFT-25082025.12 | 36.4 | |
| Qwen2.5-VL-7B2025.12 | 34.7 | |
| Qwen2.5VLSize=7B, Type=AR, Samples=>9M2025.12 | 33.4 | |
| Qwen2.5-VL-7B2025.12 | 29.4 | |
| DiffusionVLSize=7B, Type=Diff., Samples=738K2025.12 | 25 | |
| Gemma-3-27B-IT2025.12 | 24.9 | |
| Qwen2.5VLSize=3B, Type=AR, Samples=>9M2025.12 | 22.1 | |
| DiffusionVLSize=3B, Type=Diff., Samples=738K2025.12 | 20.2 | |
| LLaDA-VSize=8B, Type=Diff., Samples=16.5M2025.12 | 18.6 | |
| LLaDA-VSettings=K=-, P=0, Avg (%)=100.0%, FLOPs (Rem./Red.)=100%2026.01 | 18.38 | |
| LLaDA-FastVSettings=K=15, P=50, Avg (%)=98.7%, FLOPs (Rem./Red.)=51% (49%↓)2026.01 | 17.57 | |
| LLaDA-FastVSettings=K=15, P=70, Avg (%)=95.3%, FLOPs (Rem./Red.)=37% (63%↓)2026.01 | 16.47 | |
| Raw-FastVSettings=K=3, P=50, Avg (%)=80.0%, FLOPs (Rem./Red.)=37% (63%↓)2026.01 | 13.64 | |
| Raw-FastVSettings=K=3, P=70, Avg (%)=68.8%, FLOPs (Rem./Red.)=21% (79%↓)2026.01 | 12.14 |