Video Understanding on MSVD (Reporting Retention Metrics)
69.8AccuracyVanilla
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| VanillaBackbone=Video-LLaVA-7B, Original visual tokens=2048, Retained visual tokens=256, Reduction ratio=87.5%2026.05 | 69.8 | 3.91 | 100 | 100 | |
| LRCPBackbone=Video-LLaVA-7B, Original visual tokens=2048, Retained visual tokens=256, Reduction ratio=87.5%2026.05 | 67.7 | 3.87 | 97 | 99 | |
| VisionZipBackbone=Video-LLaVA-7B, Original visual tokens=2048, Retained visual tokens=256, Reduction ratio=87.5%2026.05 | 65.2 | 3.83 | 93.4 | 97.9 | |
| PDropBackbone=Video-LLaVA-7B, Original visual tokens=2048, Retained visual tokens=256, Reduction ratio=87.5%2026.05 | 61.5 | 3.74 | 88.1 | 95.7 |