Visual Question Answering on VQAv2 (NAR Metric)
80.8VQAv2 ScoreVanilla
Evaluation results
| Method | Links | ||
|---|---|---|---|
| VanillaBase Model=LLaVA-NEXT-7B, Token Budget=2880, Pruning Rate=0%, Loc. (Pruning Location)=-2026.05 | 80.8 | 100 | |
| HoloVBase Model=LLaVA-NEXT-7B, Token Budget=320, Pruning Rate=88.9%, Loc. (Pruning Location)=Pre2026.05 | 79.5 | 95.7 | |
| DARTBase Model=LLaVA-NEXT-7B, Token Budget=320, Pruning Rate=88.9%, Loc. (Pruning Location)=L22026.05 | 79.1 | 95.6 | |
| DiffPruneBase Model=LLaVA-NEXT-7B, Token Budget=320, Pruning Rate=88.9%, Loc. (Pruning Location)=Pre2026.05 | 78.6 | 96.1 | |
| HiREDBase Model=LLaVA-NEXT-7B, Token Budget=320, Pruning Rate=88.9%, Loc. (Pruning Location)=Pre2026.05 | 75.7 | 93.4 | |
| PDropBase Model=LLaVA-NEXT-7B, Token Budget=320, Pruning Rate=88.9%, Loc. (Pruning Location)=L8–L242026.05 | 73.5 | 90.7 | |
| FastVBase Model=LLaVA-NEXT-7B, Token Budget=320, Pruning Rate=88.9%, Loc. (Pruning Location)=L2–L322026.05 | 71.9 | 87.6 |