Multimodal Perception on MME (Granular Metrics)
2,304Total ScoreVanilla
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| VanillaBackbone=Qwen2.5-VL-7B, Token Pruning Rate=0%2026.06 | 2,304 | — | — | — | — | |
| StepPruneBackbone=Qwen2.5-VL-7B, Token Pruning Rate=66.7%2026.06 | 2,302 | — | — | — | — | |
| StepPruneBackbone=Qwen2.5-VL-7B, Token Pruning Rate=77.8%2026.06 | 2,225 | — | — | — | — | |
| StepPruneBackbone=Qwen2.5-VL-7B, Token Pruning Rate=88.9%2026.06 | 2,130 | — | — | — | — | |
| HoloVBackbone=Qwen2.5-VL-7B, Token Pruning Rate=66.7%2026.06 | 2,093 | — | — | — | — | |
| FastVBackbone=Qwen2.5-VL-7B, Token Pruning Rate=66.7%2026.06 | 2,072 | — | — | — | — | |
| HoloVBackbone=Qwen2.5-VL-7B, Token Pruning Rate=77.8%2026.06 | 2,043 | — | — | — | — | |
| FastVBackbone=Qwen2.5-VL-7B, Token Pruning Rate=77.8%2026.06 | 2,036 | — | — | — | — | |
| HoloVBackbone=Qwen2.5-VL-7B, Token Pruning Rate=88.9%2026.06 | 2,006 | — | — | — | — | |
| FastVBackbone=Qwen2.5-VL-7B, Token Pruning Rate=88.9%2026.06 | 1,940 | — | — | — | — | |
| VanillaVisual Token Budget=2880, Token Retention Ratio=100%2026.06 | 1,851 | — | — | — | — | |
| PriorTRVisual Token Budget=320, Token Retention Ratio=11.1%2026.06 | 1,745 | — | — | — | — | |
| PDropVisual Token Budget=320, Token Retention Ratio=11.1%2026.06 | 1,663 | — | — | — | — | |
| FastVVisual Token Budget=320, Token Retention Ratio=11.1%2026.06 | 1,661 | — | — | — | — | |
| PruMergeVisual Token Budget=320, Token Retention Ratio=11.1%2026.06 | 1,534 | — | — | — | — | |
| SparseVLMVisual Token Budget=320, Token Retention Ratio=11.1%2026.06 | 1,533 | — | — | — | — | |
| PNDModel=Qwen3VL-2B2026.04 | 668.33 | 190 | 160 | 158.33 | 160 | |
| regularModel=Qwen3VL-2B2026.04 | 638.33 | 190 | 155 | 143.33 | 150 | |
| PNDModel=LLaVA1.5-7B2026.04 | 621.67 | 195 | 143.33 | 123.33 | 160 | |
| PNDModel=QwenVL-7B2026.04 | 616.67 | 170 | 150 | 123.33 | 173.33 | |
| AGLAModel=LLaVA1.5-7B2026.04 | 608.67 | 195 | 138.33 | 120.33 | 155 | |
| CHASDBase Model=LLaVA-1.5 [22]2026.05 | 592.35 | 185 | 148.33 | 145 | 114.02 | |
| VAFModel=LLaVA1.5-7B2026.04 | 583.67 | 195 | 133.67 | 105 | 150 | |
| SIDBase Model=LLaVA-1.5 [22]2026.05 | 581.66 | 185 | 143.33 | 133.33 | 120 | |
| regularModel=QwenVL-7B2026.04 | 568.33 | 150 | 145 | 113.33 | 160 | |
| VCDModel=LLaVA1.5-7B2026.04 | 565.33 | 185 | 125.33 | 105 | 150 | |
| AvisCBase Model=LLaVA-1.5 [22]2026.05 | 544.18 | 189.29 | 115.76 | 119.13 | 120 | |
| GreedyBase Model=LLaVA-1.5 [22]2026.05 | 531.74 | 173.57 | 125.82 | 118.33 | 114.02 | |
| regularModel=LLaVA1.5-7B2026.04 | 531.67 | 160 | 121.67 | 105 | 145 | |
| PNDModel=InternVL2-2B2026.04 | 530 | 185 | 93.33 | 141.67 | 110 | |
| VCDBase Model=LLaVA-1.5 [22]2026.05 | 516.24 | 172.14 | 120.14 | 115 | 108.96 | |
| CHASDBase Model=InstructBLIP [5]2026.05 | 471.72 | 190 | 80 | 58.39 | 143.33 | |
| regularModel=InternVL2-2B2026.04 | 450 | 168.33 | 88.33 | 103.33 | 90 | |
| AvisCBase Model=InstructBLIP [5]2026.05 | 448.33 | 185 | 80 | 63.33 | 120 | |
| PNDModel=InstructBLIP-7B2026.04 | 423.33 | 180 | 60 | 63.33 | 120 | |
| VCDBase Model=InstructBLIP [5]2026.05 | 418 | 183 | 60 | 50 | 125 | |
| SIDBase Model=InstructBLIP [5]2026.05 | 415 | 185 | 60 | 50 | 120 | |
| GreedyBase Model=InstructBLIP [5]2026.05 | 413.33 | 155 | 65 | 55 | 138.33 | |
| AGLAModel=InstructBLIP-7B2026.04 | 406.67 | 175 | 55 | 58.33 | 118.33 | |
| VAFModel=InstructBLIP-7B2026.04 | 397.33 | 175 | 51.67 | 57.33 | 113.33 | |
| VCDModel=InstructBLIP-7B2026.04 | 387.67 | 170 | 50 | 53.33 | 114.33 | |
| regularModel=InstructBLIP-7B2026.04 | 386.66 | 170 | 50 | 53.33 | 113.33 |