High-resolution Multi-modal Understanding on V*
80.23AccuracyFull Cache
Evaluation Results
| Method | Links | |
|---|---|---|
| Full Cache2025.11 | 80.23 | |
| FlashCacheretention ratio=0.12025.11 | 80.23 | |
| FlashCacheretention ratio=0.052025.11 | 79.66 | |
| SnapKVretention ratio=0.12025.11 | 79.56 | |
| SnapKVretention ratio=0.052025.11 | 78.89 | |
| MEDAretention ratio=0.12025.11 | 78.84 | |
| StreamingLLMretention ratio=0.12025.11 | 78.77 | |
| H2Oretention ratio=0.12025.11 | 78.77 | |
| H2Oretention ratio=0.052025.11 | 78.77 | |
| StreamingLLMretention ratio=0.052025.11 | 78.65 | |
| LOOK-Mretention ratio=0.12025.11 | 78.31 | |
| LOOK-Mretention ratio=0.052025.11 | 77.78 | |
| MEDAretention ratio=0.052025.11 | 77.78 |