Visual Question Answering on VQA v2 (Score & Relative Performance)
78.5AccuracyVanilla
Evaluation results
| Method | Links | ||
|---|---|---|---|
| VanillaBase Model=LLaVA 1.5 7B, Number of Visual Tokens=5762024.12 | 78.5 | 100 | |
| VisionZipBase Model=LLaVA 1.5 7B, Number of Visual Tokens=1922024.12 | 78.5 | 100 | |
| VisionZipBase Model=LLaVA 1.5 7B, Number of Visual Tokens=1282024.12 | 77.8 | 99.1 | |
| VisionZipBase Model=LLaVA 1.5 7B, Number of Visual Tokens=642024.12 | 76 | 96.8 |