Visual Question Answering on VQA v2 (VQAv2 Score and Average Metric)
79.2VQAv2 ScoreLLaVA-SP
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| LLaVA-SPLLM=Vicuna-7B, Resolution=3362026.07 | 79.2 | — | |
| Qwen-VLLLM=Qwen-7B, Resolution=4482026.07 | 78.8 | — | |
| LLaVA-1.5-7BVision Token=5762026.06 | 78.5 | 100 | |
| LLaVA-1.5LLM=Vicuna-7B, Resolution=3362026.07 | 78.5 | — | |
| LLaVA-1.5LLM=Vicuna-7B, Resolution=3362026.07 | 78.4 | — | |
| Qwen-VL-ChatLLM=Qwen-7B, Resolution=4482026.07 | 78.2 | — | |
| MS-ResamplerVision Token=1442026.06 | 78 | 100.43 | |
| DARTVision Token=1922026.06 | 76.7 | 97.98 | |
| TrustLLaVALLM=Vicuna-7B, Resolution=3362026.07 | 76.3 | — | |
| MustDropVision Token=1922026.06 | 76 | 96.26 | |
| SparseVLMVision Token=1922026.06 | 75.6 | 95.79 | |
| PDropVision Token=1922026.06 | 75.1 | 96.19 | |
| HiREDVision Token=1922026.06 | 74.9 | 89.06 | |
| TrustLLaVA (MLP)LLM=Vicuna-7B, Resolution=3362026.07 | 72.8 | — | |
| LLaVA-PruMergeVision Token=1922026.06 | 70.6 | 88.92 | |
| FastVVision Token=1922026.06 | 67.1 | 89.58 |