Visual Question Answering on VQA text
83.2AccuracyQ2.5VL-7B
Evaluation Results
| Method | Links | |
|---|---|---|
| Q2.5VL-7BTokens Retained Percentage=100%2025.03 | 83.2 | |
| TwigVLM++Tokens Retained Percentage=33.3%2025.03 | 83.2 | |
| TwigVLMTokens Retained Percentage=33.3%2025.03 | 82.6 | |
| TwigVLM++Tokens Retained Percentage=22.2%2025.03 | 82.5 | |
| VanillaBackbone=QWEN-2.5-VL 7B, Average Tokens=100%2026.02 | 82.2 | |
| FastVTokens Retained Percentage=33.3%2025.03 | 81.5 | |
| TwigVLMTokens Retained Percentage=22.2%2025.03 | 81.4 | |
| TwigVLM++Tokens Retained Percentage=11.1%2025.03 | 79.5 | |
| FastVTokens Retained Percentage=22.2%2025.03 | 79.2 | |
| NuwaBackbone=QWEN-2.5-VL 7B, Average Tokens=75%2026.02 | 77.35 | |
| VisionZipTokens Retained Percentage=33.3%2025.03 | 76.3 | |
| NuwaBackbone=QWEN-2.5-VL 7B, Average Tokens=50%2026.02 | 76.03 | |
| NuwaBackbone=QWEN-2.5-VL 7B, Average Tokens=25%2026.02 | 73.81 | |
| TwigVLMTokens Retained Percentage=11.1%2025.03 | 73 | |
| VisionZipTokens Retained Percentage=22.2%2025.03 | 71.6 | |
| FastVTokens Retained Percentage=11.1%2025.03 | 62.9 | |
| VisionZipTokens Retained Percentage=11.1%2025.03 | 58.7 | |
| LLaVA-1.5-7BRetained Tokens=5762025.09 | 58.2 | |
| VanillaBase Model=LLaVA-1.5-7B, Retained Tokens=576, Reduction Rate=0%2026.02 | 58.2 | |
| ZOO-PruneRetained Tokens=1282025.09 | 57.87 | |
| VisionZipRetained Tokens=1922025.09 | 57.3 | |
| ZOO-PruneRetained Tokens=1922025.09 | 57.3 | |
| VisionZipBase Model=LLaVA-1.5-7B, Retained Tokens=192, Reduction Rate=66.7%2026.02 | 57.3 | |
| DivPruneRetained Tokens=1922025.09 | 56.97 | |
| VisionZipRetained Tokens=1282025.09 | 56.8 | |
| VisionZipBase Model=LLaVA-1.5-7B, Retained Tokens=128, Reduction Rate=77.8%2026.02 | 56.8 | |
| PDropBase Model=LLaVA-1.5-7B, Retained Tokens=128, Reduction Rate=77.8%2026.02 | 56.6 | |
| PDropBase Model=LLaVA-1.5-7B, Retained Tokens=192, Reduction Rate=66.7%2026.02 | 56.5 | |
| PyramidDropRetained Tokens=1922025.09 | 56.1 | |
| SparseVLMRetained Tokens=1922025.09 | 56.1 | |
| SparseVLMBase Model=LLaVA-1.5-7B, Retained Tokens=192, Reduction Rate=66.7%2026.02 | 56.1 | |
| DivPruneRetained Tokens=1282025.09 | 56.06 | |
| VisionZipRetained Tokens=642025.09 | 55.5 | |
| VisionZipBase Model=LLaVA-1.5-7B, Retained Tokens=64, Reduction Rate=88.9%2026.02 | 55.5 | |
| ZOO-PruneRetained Tokens=642025.09 | 55.35 | |
| PyramidDropRetained Tokens=1282025.09 | 55.1 | |
| SparseVLMRetained Tokens=1282025.09 | 54.9 | |
| SparseVLMBase Model=LLaVA-1.5-7B, Retained Tokens=128, Reduction Rate=77.8%2026.02 | 54.9 | |
| DivPruneRetained Tokens=642025.09 | 54.69 | |
| ApETBase Model=LLaVA-1.5-7B, Retained Tokens=192, Reduction Rate=66.7%2026.02 | 54.4 | |
| ApETBase Model=LLaVA-1.5-7B, Retained Tokens=128, Reduction Rate=77.8%2026.02 | 53.9 | |
| ApETBase Model=LLaVA-1.5-7B, Retained Tokens=64, Reduction Rate=88.9%2026.02 | 53 | |
| FastVRetained Tokens=1922025.09 | 52.5 | |
| FastVBase Model=LLaVA-1.5-7B, Retained Tokens=192, Reduction Rate=66.7%2026.02 | 52.5 | |
| ToMeRetained Tokens=1922025.09 | 52.1 | |
| ToMeBase Model=LLaVA-1.5-7B, Retained Tokens=192, Reduction Rate=66.7%2026.02 | 52.1 | |
| SparseVLMRetained Tokens=642025.09 | 51.8 | |
| SparseVLMBase Model=LLaVA-1.5-7B, Retained Tokens=64, Reduction Rate=88.9%2026.02 | 51.8 | |
| FastVRetained Tokens=1282025.09 | 50.6 | |
| FastVBase Model=LLaVA-1.5-7B, Retained Tokens=128, Reduction Rate=77.8%2026.02 | 50.6 | |
| PDropBase Model=LLaVA-1.5-7B, Retained Tokens=64, Reduction Rate=88.9%2026.02 | 50.6 | |
| ToMeRetained Tokens=1282025.09 | 49.1 | |
| ToMeBase Model=LLaVA-1.5-7B, Retained Tokens=128, Reduction Rate=77.8%2026.02 | 49.1 | |
| FastVRetained Tokens=642025.09 | 47.8 | |
| FastVBase Model=LLaVA-1.5-7B, Retained Tokens=64, Reduction Rate=88.9%2026.02 | 47.8 | |
| HiREDRetained Tokens=1922025.09 | 47.4 | |
| HiREDRetained Tokens=1282025.09 | 46.1 | |
| PyramidDropRetained Tokens=642025.09 | 45.9 | |
| ToMeRetained Tokens=642025.09 | 45.3 | |
| ToMeBase Model=LLaVA-1.5-7B, Retained Tokens=64, Reduction Rate=88.9%2026.02 | 45.3 | |
| HiREDRetained Tokens=642025.09 | 44.2 |