Aggregated Multimodal Evaluation on Multimodal Benchmark Average
100Average ScoreQwen-2.5-VL-7B
Evaluation Results
| Method | Links | |
|---|---|---|
| Qwen-2.5-VL-7BResolution Strategy=Dynamic, Token Retention Ratio=100%2025.08 | 100 | |
| Qwen-2.5-VL-7BResolution Strategy=Fixed, Token Retention Ratio=100%2025.08 | 99.3 | |
| MMTokResolution Strategy=Dynamic, Token Retention Ratio=20%2025.08 | 94.6 | |
| VisionZipResolution Strategy=Dynamic, Token Retention Ratio=20%2025.08 | 94.2 | |
| DivPruneResolution Strategy=Dynamic, Token Retention Ratio=20%2025.08 | 91.5 | |
| MMTokResolution Strategy=Dynamic, Token Retention Ratio=10%2025.08 | 88.5 | |
| VisionZipResolution Strategy=Dynamic, Token Retention Ratio=10%2025.08 | 87.5 | |
| DivPruneResolution Strategy=Dynamic, Token Retention Ratio=10%2025.08 | 84.7 | |
| MMTokResolution Strategy=Dynamic, Token Retention Ratio=5%2025.08 | 79 | |
| DivPruneResolution Strategy=Dynamic, Token Retention Ratio=5%2025.08 | 76.3 | |
| VisionZipResolution Strategy=Dynamic, Token Retention Ratio=5%2025.08 | 75.4 | |
| Qwen-2.5-VL-7BToken Retention Ratio=0%2025.08 | 33.8 |