Image-text understanding on MME
1,702.2MME ScoreVisPruner
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| VisPrunerToken Budget=512, Reduction Percentage=60.5%2026.03 | 1,702.2 | 94 | |
| Qwen2.5-VL 7BToken Budget=All 1296 Tokens, Reduction Percentage=0%2026.03 | 1,685.9 | 100 | |
| PromPruneToken Budget=256, Reduction Percentage=80.2%2026.03 | 1,682.4 | 92.6 | |
| DivPruneToken Budget=256, Reduction Percentage=80.2%2026.03 | 1,681.2 | 91.1 | |
| PromPruneToken Budget=512, Reduction Percentage=60.5%2026.03 | 1,678.6 | 94.8 | |
| DivPruneToken Budget=512, Reduction Percentage=60.5%2026.03 | 1,676 | 94.2 | |
| VisPrunerToken Budget=256, Reduction Percentage=80.2%2026.03 | 1,649.5 | 91.7 | |
| PromPruneToken Budget=128, Reduction Percentage=90.1%2026.03 | 1,624.2 | 87.9 | |
| DivPruneToken Budget=128, Reduction Percentage=90.1%2026.03 | 1,578.5 | 86.5 | |
| VisPrunerToken Budget=128, Reduction Percentage=90.1%2026.03 | 1,564.2 | 86.2 | |
| LLaVA-NeXT 7BToken Budget=2880, Base Model=LLaVA-NeXT 7B2026.03 | 1,507 | 100 | |
| PromPruneToken Budget=640, Base Model=LLaVA-NeXT 7B2026.03 | 1,489.8 | 99.6 | |
| VisPrunerToken Budget=640, Base Model=LLaVA-NeXT 7B2026.03 | 1,486.1 | 99.2 | |
| DivPruneToken Budget=640, Base Model=LLaVA-NeXT 7B2026.03 | 1,482.6 | 99.6 | |
| PromPruneToken Budget=320, Base Model=LLaVA-NeXT 7B2026.03 | 1,471 | 97.9 | |
| DivPruneToken Budget=320, Base Model=LLaVA-NeXT 7B2026.03 | 1,438.9 | 97.7 | |
| VisPrunerToken Budget=320, Base Model=LLaVA-NeXT 7B2026.03 | 1,435.8 | 95.9 | |
| PromPruneToken Budget=160, Base Model=LLaVA-NeXT 7B2026.03 | 1,361.3 | 95.2 | |
| DivPruneToken Budget=160, Base Model=LLaVA-NeXT 7B2026.03 | 1,349 | 95 | |
| VisPrunerToken Budget=160, Base Model=LLaVA-NeXT 7B2026.03 | 1,315.4 | 90.6 |