Multimodal Reasoning on RealWorldQA (Accuracy, Token Length)
481.8Generated Token LengthLOOK-M
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| LOOK-MRetention level=≤ 10%2026.05 | 481.8 | 67.84 | |
| FastV-VPRetention level=≤ 5%2026.05 | 538 | 72.54 | |
| FastV-VPRetention level=≤ 10%2026.05 | 564 | 73.2 | |
| LOOK-MRetention level=≤ 5%2026.05 | 573 | 68.37 | |
| FastVRetention level=≤ 10%2026.05 | 575.8 | 55.95 | |
| FastVRetention level=≤ 5%2026.05 | 613.3 | 54.12 | |
| VisionzipRetention level=≤ 5%2026.05 | 671.2 | 52.02 | |
| BaselineRetention level=100%2026.05 | 678 | 72.81 | |
| VisionzipRetention level=≤ 10%2026.05 | 726 | 59.08 |