Multimodal Reasoning on LMMs-Eval Average
70.83Average AccuracyInternVL2-8B
Evaluation Results
| Method | Links | |
|---|---|---|
| InternVL2-8BReduction Method=None (Base), Memory cost=1, TFLOPs=24.1, Time cost=1.5182024.11 | 70.83 | |
| G-SearchBackbone=InternVL2-8B, Memory cost=0.424, TFLOPs=12.24, Time cost=0.862024.11 | 70.1 | |
| FastVBackbone=InternVL2-8B, Filtering Ratio (R)=50%, Memory cost=0.531, TFLOPs=14.58, Time cost=0.9982024.11 | 69.42 | |
| PDropBackbone=InternVL2-8B, Memory cost=0.469, TFLOPs=13.13, Time cost=0.9152024.11 | 69.19 | |
| InternVL2-4BReduction Method=None (Base), Memory cost=1, TFLOPs=13.97, Time cost=0.9692024.11 | 68.16 | |
| G-SearchBackbone=InternVL2-4B, Memory cost=0.488, TFLOPs=8.69, Time cost=0.6452024.11 | 67.65 | |
| PDropBackbone=InternVL2-4B, Memory cost=0.469, TFLOPs=8.35, Time cost=0.6272024.11 | 66.94 | |
| FastVBackbone=InternVL2-4B, Filtering Ratio (R)=50%, Memory cost=0.531, TFLOPs=9.01, Time cost=0.6522024.11 | 66.19 | |
| InternVL2-2BReduction Method=None (Base), Memory cost=1, TFLOPs=8.1, Time cost=0.5982024.11 | 61.94 | |
| G-SearchBackbone=InternVL2-2B, Memory cost=0.532, TFLOPs=5.64, Time cost=0.4442024.11 | 61.22 | |
| FastVBackbone=InternVL2-2B, Filtering Ratio (R)=50%, Memory cost=0.542, TFLOPs=5.85, Time cost=0.4512024.11 | 59.91 | |
| InternVL2-1BReduction Method=None (Base), Memory cost=1, TFLOPs=4.62, Time cost=0.3842024.11 | 59.85 | |
| G-SearchBackbone=InternVL2-1B, Memory cost=0.527, TFLOPs=3.84, Time cost=0.3332024.11 | 59.19 | |
| PDropBackbone=InternVL2-2B, Memory cost=0.583, TFLOPs=5.93, Time cost=0.4522024.11 | 58.98 | |
| FastVBackbone=InternVL2-1B, Filtering Ratio (R)=50%, Memory cost=0.542, TFLOPs=3.91, Time cost=0.3422024.11 | 54.85 | |
| PDropBackbone=InternVL2-1B, Memory cost=0.583, TFLOPs=3.88, Time cost=0.3362024.11 | 53.7 | |
| VTWBackbone=InternVL2-8B, Memory cost=0.5, TFLOPs=13.71, Time cost=0.9272024.11 | 52.51 | |
| VTWBackbone=InternVL2-4B, Memory cost=0.5, TFLOPs=8.72, Time cost=0.6492024.11 | 49.01 | |
| LLaVA-1.5-7BReduction Method=None (Base), Memory cost=1, TFLOPs=9.18, Time cost=0.6252024.11 | 48.97 | |
| G-SearchBackbone=LLaVA-1.5-7B, Memory cost=0.34, TFLOPs=3.95, Time cost=0.3012024.11 | 48.77 | |
| PDropBackbone=LLaVA-1.5-7B, Memory cost=0.469, TFLOPs=5.47, Time cost=0.3812024.11 | 48.7 | |
| FastVBackbone=LLaVA-1.5-7B, Filtering Ratio (R)=50%, Memory cost=0.531, TFLOPs=5.49, Time cost=0.3872024.11 | 48.7 | |
| VTWBackbone=LLaVA-1.5-7B, Memory cost=0.5, TFLOPs=5.19, Time cost=0.3852024.11 | 44.32 | |
| VTWBackbone=InternVL2-1B, Memory cost=0.5, TFLOPs=3.73, Time cost=0.3312024.11 | 41.13 | |
| VTWBackbone=InternVL2-2B, Memory cost=0.5, TFLOPs=5.5, Time cost=0.4392024.11 | 37.84 |