Video Understanding on TGIF, MSVD, MSRVTT, ActNet Averaged
100Relative AccuracyVideo-LLaVA
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Video-LLaVA2026.05 | 100 | — | |
| OccamToken2026.05 | 97.7 | — | |
| VisionZip2026.05 | 93.2 | — | |
| SparseVLM2026.05 | 86.5 | — | |
| CLSE-MEvaluator=Gemini-2.5-Flash, Pruning Strategy=w/ Token Merging, Retained Tokens=1942026.06 | 42.3 | 2.11 | |
| SparseVLMEvaluator=Gemini-2.5-Flash, Pruning Strategy=w/ Token Merging, Retained Tokens=1942026.06 | 40.8 | 2.09 | |
| Video-LLaVA-7BEvaluator=Gemini-2.5-Flash, Pruning Strategy=None, Retained Tokens=20482026.06 | 40.4 | 2.04 | |
| CLSEEvaluator=Gemini-2.5-Flash, Pruning Strategy=Hard Pruning, Retained Tokens=1942026.06 | 39.2 | 1.95 | |
| Video-LLaVA-7BEvaluator=GPT-4o-mini, Pruning Strategy=None, Retained Tokens=20482026.06 | 38.7 | 2.35 | |
| CLSE-MEvaluator=GPT-4o-mini, Pruning Strategy=w/ Token Merging, Retained Tokens=1942026.06 | 38.7 | 2.29 | |
| SparseVLMEvaluator=GPT-4o-mini, Pruning Strategy=w/ Token Merging, Retained Tokens=1942026.06 | 38.1 | 2.27 | |
| FiCoCo-VEvaluator=Gemini-2.5-Flash, Pruning Strategy=Hard Pruning, Retained Tokens=1942026.06 | 36.7 | 1.79 | |
| CLSEEvaluator=GPT-4o-mini, Pruning Strategy=Hard Pruning, Retained Tokens=1942026.06 | 35.7 | 2.19 | |
| PDropEvaluator=Gemini-2.5-Flash, Pruning Strategy=Hard Pruning, Retained Tokens=1942026.06 | 33 | 1.62 | |
| FiCoCo-VEvaluator=GPT-4o-mini, Pruning Strategy=Hard Pruning, Retained Tokens=1942026.06 | 31.9 | 2.13 | |
| PDropEvaluator=GPT-4o-mini, Pruning Strategy=Hard Pruning, Retained Tokens=1942026.06 | 27.3 | 1.82 | |
| FastVEvaluator=Gemini-2.5-Flash, Pruning Strategy=Hard Pruning, Retained Tokens=1942026.06 | 26.6 | 1.36 | |
| FastVEvaluator=GPT-4o-mini, Pruning Strategy=Hard Pruning, Retained Tokens=1942026.06 | 26.1 | 1.88 |