Multimodal Evaluation on Combined Benchmark Suite (GQA, MMB, MME, VQA-T, SQA-I, VQA-v2)
100Relative AccuracyLLaVA-NeXT-7B
Evaluation Results
| Method | Links | |
|---|---|---|
| LLaVA-NeXT-7BRetained Tokens=2880, Base Model=LLaVA-NeXT-7B2025.03 | 100 | |
| LLaVA-NeXT-7BRetained Tokens=2880, Base Model=LLaVA-NeXT-7B2025.03 | 100 | |
| TwigVLMRetained Tokens=640, Base Model=LLaVA-NeXT-7B2025.03 | 99 | |
| TwigVLM++Retained Tokens=640, Base Model=LLaVA-NeXT-7B2025.03 | 99 | |
| TwigVLMRetained Tokens=640, Base Model=LLaVA-NeXT-7B2025.03 | 99 | |
| TwigVLM++Retained Tokens=640, Base Model=LLaVA-NeXT-7B2025.03 | 99 | |
| FastVRetained Tokens=640, Base Model=LLaVA-NeXT-7B2025.03 | 97.6 | |
| FastVRetained Tokens=640, Base Model=LLaVA-NeXT-7B2025.03 | 97.6 | |
| VisionZip‡Retained Tokens=640, Base Model=LLaVA-NeXT-7B2025.03 | 97.4 | |
| VisionZip‡Retained Tokens=640, Base Model=LLaVA-NeXT-7B2025.03 | 97.4 | |
| VisionZipRetained Tokens=640, Base Model=LLaVA-NeXT-7B2025.03 | 97.1 | |
| TwigVLM++Retained Tokens=320, Base Model=LLaVA-NeXT-7B2025.03 | 97.1 | |
| VisionZipRetained Tokens=640, Base Model=LLaVA-NeXT-7B2025.03 | 97.1 | |
| TwigVLM++Retained Tokens=320, Base Model=LLaVA-NeXT-7B2025.03 | 97.1 | |
| TwigVLMRetained Tokens=320, Base Model=LLaVA-NeXT-7B2025.03 | 96.2 | |
| TwigVLMRetained Tokens=320, Base Model=LLaVA-NeXT-7B2025.03 | 96.2 | |
| VisionZip‡Retained Tokens=320, Base Model=LLaVA-NeXT-7B2025.03 | 95.8 | |
| VisionZip‡Retained Tokens=320, Base Model=LLaVA-NeXT-7B2025.03 | 95.8 | |
| SparseVLMRetained Tokens=640, Base Model=LLaVA-NeXT-7B2025.03 | 95.4 | |
| SparseVLMRetained Tokens=640, Base Model=LLaVA-NeXT-7B2025.03 | 95.4 | |
| VisionZipRetained Tokens=320, Base Model=LLaVA-NeXT-7B2025.03 | 93.8 | |
| VisionZipRetained Tokens=320, Base Model=LLaVA-NeXT-7B2025.03 | 93.8 | |
| MustDropRetained Tokens=320, Base Model=LLaVA-NeXT-7B2025.03 | 92.6 | |
| MustDropRetained Tokens=320, Base Model=LLaVA-NeXT-7B2025.03 | 92.6 | |
| SparseVLMRetained Tokens=320, Base Model=LLaVA-NeXT-7B2025.03 | 92.3 | |
| SparseVLMRetained Tokens=320, Base Model=LLaVA-NeXT-7B2025.03 | 92.3 | |
| FastVRetained Tokens=320, Base Model=LLaVA-NeXT-7B2025.03 | 88.2 | |
| FastVRetained Tokens=320, Base Model=LLaVA-NeXT-7B2025.03 | 88.2 |