Visual Question Answering on GQA v1 (test)
64.2GQA AccuracyLLaVA-Next-7B
Evaluation results
| Method | Links | ||
|---|---|---|---|
| LLaVA-Next-7BBase Model=LLaVA-Next-7B, Reduction Ratio=100%, Retained Tokens=28802026.06 | 64.2 | 100 | |
| HoloV + RESTOREBase Model=LLaVA-Next-7B, Reduction Ratio=11.1%, Retained Tokens=3202026.06 | 61 | 92 | |
| VisPruner + RESTOREBase Model=LLaVA-Next-7B, Reduction Ratio=11.1%, Retained Tokens=3202026.06 | 60.9 | 91.9 | |
| DivPrune + RESTOREBase Model=LLaVA-Next-7B, Reduction Ratio=11.1%, Retained Tokens=3202026.06 | 60.7 | 90.2 | |
| Visionzip + RESTOREBase Model=LLaVA-Next-7B, Reduction Ratio=11.1%, Retained Tokens=3202026.06 | 60.6 | 90.5 | |
| DivPruneBase Model=LLaVA-Next-7B, Reduction Ratio=11.1%, Retained Tokens=3202026.06 | 60.2 | 88.8 | |
| HoloVBase Model=LLaVA-Next-7B, Reduction Ratio=11.1%, Retained Tokens=3202026.06 | 59.6 | 87.5 | |
| VisPrunerBase Model=LLaVA-Next-7B, Reduction Ratio=11.1%, Retained Tokens=3202026.06 | 59.3 | 90.8 | |
| HoloV + RESTOREBase Model=LLaVA-Next-7B, Reduction Ratio=5.5%, Retained Tokens=1602026.06 | 59.2 | 89 | |
| VisionzipBase Model=LLaVA-Next-7B, Reduction Ratio=11.1%, Retained Tokens=3202026.06 | 59 | 88.9 | |
| DivPrune + RESTOREBase Model=LLaVA-Next-7B, Reduction Ratio=5.5%, Retained Tokens=1602026.06 | 59 | 87.2 | |
| VisPruner + RESTOREBase Model=LLaVA-Next-7B, Reduction Ratio=5.5%, Retained Tokens=1602026.06 | 58.8 | 88.5 | |
| Visionzip + RESTOREBase Model=LLaVA-Next-7B, Reduction Ratio=5.5%, Retained Tokens=1602026.06 | 58.5 | 87.9 | |
| DivPruneBase Model=LLaVA-Next-7B, Reduction Ratio=5.5%, Retained Tokens=1602026.06 | 58.4 | 86.3 | |
| HoloVBase Model=LLaVA-Next-7B, Reduction Ratio=5.5%, Retained Tokens=1602026.06 | 57.2 | 86.3 | |
| VisPrunerBase Model=LLaVA-Next-7B, Reduction Ratio=5.5%, Retained Tokens=1602026.06 | 57 | 86.7 | |
| VisionzipBase Model=LLaVA-Next-7B, Reduction Ratio=5.5%, Retained Tokens=1602026.06 | 56.2 | 84.9 |