Visual Question Answering on GQA (GQA, Rel.)
61.9GQA ScoreVanilla
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| VanillaToken Budget (Retained Tokens)=576, Reduction Ratio (↓ %)=0%, Base Model (LLaVA-1.5-7B)=LLaVA-1.5-7B2026.05 | 61.9 | 100 | |
| PACTToken Budget (Retained Tokens)=192, Reduction Ratio (↓ %)=66.7%, Base Model (LLaVA-1.5-7B)=LLaVA-1.5-7B2026.05 | 60.7 | 98.1 | |
| SPprunerToken Budget (Retained Tokens)=192, Reduction Ratio (↓ %)=66.7%, Base Model (LLaVA-1.5-7B)=LLaVA-1.5-7B2026.05 | 60.5 | 99.4 | |
| PACTToken Budget (Retained Tokens)=128, Reduction Ratio (↓ %)=77.8%, Base Model (LLaVA-1.5-7B)=LLaVA-1.5-7B2026.05 | 60.5 | 96.9 | |
| DivPruneToken Budget (Retained Tokens)=192, Reduction Ratio (↓ %)=66.7%, Base Model (LLaVA-1.5-7B)=LLaVA-1.5-7B2026.05 | 60.1 | 98.8 | |
| DARTToken Budget (Retained Tokens)=192, Reduction Ratio (↓ %)=66.7%, Base Model (LLaVA-1.5-7B)=LLaVA-1.5-7B2026.05 | 60 | 99 | |
| PACTToken Budget (Retained Tokens)=64, Reduction Ratio (↓ %)=88.9%, Base Model (LLaVA-1.5-7B)=LLaVA-1.5-7B2026.05 | 59.5 | 94 | |
| PyramidDropToken Budget (Retained Tokens)=192, Reduction Ratio (↓ %)=66.7%, Base Model (LLaVA-1.5-7B)=LLaVA-1.5-7B2026.05 | 59.3 | 98.7 | |
| VisionZipToken Budget (Retained Tokens)=192, Reduction Ratio (↓ %)=66.7%, Base Model (LLaVA-1.5-7B)=LLaVA-1.5-7B2026.05 | 59.3 | 99 | |
| DivPruneToken Budget (Retained Tokens)=128, Reduction Ratio (↓ %)=77.8%, Base Model (LLaVA-1.5-7B)=LLaVA-1.5-7B2026.05 | 59.3 | 97.4 | |
| SPprunerToken Budget (Retained Tokens)=128, Reduction Ratio (↓ %)=77.8%, Base Model (LLaVA-1.5-7B)=LLaVA-1.5-7B2026.05 | 59.2 | 98.5 | |
| DARTToken Budget (Retained Tokens)=128, Reduction Ratio (↓ %)=77.8%, Base Model (LLaVA-1.5-7B)=LLaVA-1.5-7B2026.05 | 58.2 | 97.2 | |
| VisionZipToken Budget (Retained Tokens)=128, Reduction Ratio (↓ %)=77.8%, Base Model (LLaVA-1.5-7B)=LLaVA-1.5-7B2026.05 | 57.7 | 97.6 | |
| DivPruneToken Budget (Retained Tokens)=64, Reduction Ratio (↓ %)=88.9%, Base Model (LLaVA-1.5-7B)=LLaVA-1.5-7B2026.05 | 57.7 | 94.4 | |
| PyramidDropToken Budget (Retained Tokens)=128, Reduction Ratio (↓ %)=77.8%, Base Model (LLaVA-1.5-7B)=LLaVA-1.5-7B2026.05 | 57.6 | 96.5 | |
| SPprunerToken Budget (Retained Tokens)=64, Reduction Ratio (↓ %)=88.9%, Base Model (LLaVA-1.5-7B)=LLaVA-1.5-7B2026.05 | 56.3 | 95.2 | |
| DARTToken Budget (Retained Tokens)=64, Reduction Ratio (↓ %)=88.9%, Base Model (LLaVA-1.5-7B)=LLaVA-1.5-7B2026.05 | 56 | 93.4 | |
| VisionZipToken Budget (Retained Tokens)=64, Reduction Ratio (↓ %)=88.9%, Base Model (LLaVA-1.5-7B)=LLaVA-1.5-7B2026.05 | 55.1 | 90.4 | |
| PyramidDropToken Budget (Retained Tokens)=64, Reduction Ratio (↓ %)=88.9%, Base Model (LLaVA-1.5-7B)=LLaVA-1.5-7B2026.05 | 54.6 | 88.6 | |
| FastVToken Budget (Retained Tokens)=192, Reduction Ratio (↓ %)=66.7%, Base Model (LLaVA-1.5-7B)=LLaVA-1.5-7B2026.05 | 52.7 | 90 | |
| FastVToken Budget (Retained Tokens)=128, Reduction Ratio (↓ %)=77.8%, Base Model (LLaVA-1.5-7B)=LLaVA-1.5-7B2026.05 | 49.6 | 82.9 | |
| FastVToken Budget (Retained Tokens)=64, Reduction Ratio (↓ %)=88.9%, Base Model (LLaVA-1.5-7B)=LLaVA-1.5-7B2026.05 | 46.1 | 74 |