Science Question Answering on SQA (Score, Relative Performance)
70.15ScoreVanilla
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| VanillaModel=LLaVA-NeXT, TFLOPs=100%2026.06 | 70.15 | 100 | |
| VTW (K=16)Model=LLaVA-NeXT, TFLOPs=51%2026.06 | 70 | 86.5 | |
| Ours (N=20)Model=LLaVA-NeXT, TFLOPs=67%2026.06 | 69.96 | 98.8 | |
| V2DropModel=LLaVA-NeXT, TFLOPs=70%2026.06 | 69.92 | 98.4 | |
| VSkip+ (N=20)Model=LLaVA-NeXT, TFLOPs=74%2026.06 | 69.83 | 98.2 | |
| APETModel=LLaVA-NeXT, TFLOPs=70%2026.06 | 69.71 | 99.4 | |
| OC-VTPTokens=640, Retaining Ratio=22.2%, Backbone=LLaVA-NeXT2025.11 | 69.3 | 103 | |
| OC-VTPTokens=640, Retaining Ratio=22.2%, Individual Training=true, Backbone=LLaVA-NeXT2025.11 | 69.3 | 103 | |
| ShortV (N=20)Model=LLaVA-NeXT, TFLOPs=51%2026.06 | 69.26 | 96 | |
| OC-VTPTokens=160, Retaining Ratio=5.6%, Backbone=LLaVA-NeXT2025.11 | 69.2 | 103 | |
| VisionZipTokens=160, Retaining Ratio=5.6%, Backbone=LLaVA-NeXT2025.11 | 69 | 102 | |
| OC-VTPTokens=160, Retaining Ratio=5.6%, Individual Training=true, Backbone=LLaVA-NeXT2025.11 | 69 | 102 | |
| HiPruneTokens=160, Retaining Ratio=5.6%, Backbone=LLaVA-NeXT2025.11 | 68.7 | 102 | |
| VisionZipTokens=640, Retaining Ratio=22.2%, Backbone=LLaVA-NeXT2025.11 | 68.1 | 101 | |
| HiPruneTokens=640, Retaining Ratio=22.2%, Backbone=LLaVA-NeXT2025.11 | 68 | 101 | |
| VanillaTokens=2880, Retaining Ratio=100%, Backbone=LLaVA-NeXT2025.11 | 67.5 | 100 | |
| OC-VTPTokens=320, Retaining Ratio=11.1%, Backbone=LLaVA-NeXT2025.11 | 67.4 | 99.9 | |
| VisionZipTokens=320, Retaining Ratio=11.1%, Backbone=LLaVA-NeXT2025.11 | 67.3 | 99.7 | |
| HiPruneTokens=320, Retaining Ratio=11.1%, Backbone=LLaVA-NeXT2025.11 | 67.3 | 99.7 | |
| OC-VTPTokens=320, Retaining Ratio=11.1%, Individual Training=true, Backbone=LLaVA-NeXT2025.11 | 67.3 | 99.7 |