Visual Question Answering on GQA (accuracy, relative accuracy)
61.9AccuracyVanilla
Evaluation results
| Method | Links | ||
|---|---|---|---|
| VanillaRetained Tokens=576, Pruning Ratio=100%, Training Protocol=Zero-shot2025.11 | 61.9 | 100 | |
| OC-VTPRetained Tokens=192, Pruning Ratio=33.3%, Training Protocol=Benchmark-specific training2025.11 | 59.6 | 96.3 | |
| OC-VTPRetained Tokens=192, Pruning Ratio=33.3%, Training Protocol=Zero-shot2025.11 | 59.4 | 95.9 | |
| VisionZipRetained Tokens=192, Pruning Ratio=33.3%, Training Protocol=Zero-shot2025.11 | 59.3 | 95.8 | |
| HiPruneRetained Tokens=192, Pruning Ratio=33.3%, Training Protocol=Zero-shot2025.11 | 59.2 | 95.6 | |
| OC-VTPRetained Tokens=128, Pruning Ratio=22.2%, Training Protocol=Benchmark-specific training2025.11 | 58.4 | 94.3 | |
| SparseVLMRetained Tokens=192, Pruning Ratio=33.3%, Training Protocol=Zero-shot2025.11 | 57.6 | 93.1 | |
| VisionZipRetained Tokens=128, Pruning Ratio=22.2%, Training Protocol=Zero-shot2025.11 | 57.6 | 93.1 | |
| OC-VTPRetained Tokens=128, Pruning Ratio=22.2%, Training Protocol=Zero-shot2025.11 | 57.5 | 92.9 | |
| PyramidDropRetained Tokens=192, Pruning Ratio=33.3%, Training Protocol=Zero-shot2025.11 | 57.3 | 92.6 | |
| HiPruneRetained Tokens=128, Pruning Ratio=22.2%, Training Protocol=Zero-shot2025.11 | 57.3 | 92.6 | |
| PyramidDropRetained Tokens=128, Pruning Ratio=22.2%, Training Protocol=Zero-shot2025.11 | 57.1 | 92.2 | |
| OC-VTPRetained Tokens=64, Pruning Ratio=11.1%, Training Protocol=Benchmark-specific training2025.11 | 56.2 | 90.8 | |
| SparseVLMRetained Tokens=128, Pruning Ratio=22.2%, Training Protocol=Zero-shot2025.11 | 56 | 90.5 | |
| OC-VTPRetained Tokens=50.7, Pruning Ratio=8.8%, Training Protocol=Zero-shot2025.11 | 55.3 | 89.3 | |
| VisionZipRetained Tokens=64, Pruning Ratio=11.1%, Training Protocol=Zero-shot2025.11 | 55.1 | 89 | |
| OC-VTPRetained Tokens=64, Pruning Ratio=11.1%, Training Protocol=Zero-shot2025.11 | 54.6 | 88.1 | |
| HiPruneRetained Tokens=64, Pruning Ratio=11.1%, Training Protocol=Zero-shot2025.11 | 53.6 | 86.6 | |
| FastVECRetained Tokens=192, Pruning Ratio=33.3%, Training Protocol=Zero-shot2025.11 | 52.7 | 85.1 | |
| SparseVLMRetained Tokens=64, Pruning Ratio=11.1%, Training Protocol=Zero-shot2025.11 | 52.7 | 85.1 | |
| FastVECRetained Tokens=128, Pruning Ratio=22.2%, Training Protocol=Zero-shot2025.11 | 49.6 | 80.1 | |
| PyramidDropRetained Tokens=64, Pruning Ratio=11.1%, Training Protocol=Zero-shot2025.11 | 47.5 | 76.7 | |
| FastVECRetained Tokens=64, Pruning Ratio=11.1%, Training Protocol=Zero-shot2025.11 | 46.1 | 74.5 |