Long-context Video Understanding on VideoMME (test)
69.3ScoreQwen3-VL-4B
Evaluation results
| Method | Links | |
|---|---|---|
| Qwen3-VL-4BModel=Qwen3-VL-4B, Pruning Strength=None2026.03 | 69.3 | |
| InternVL3.5-8BModel=InternVL3.5-8B, Pruning Strength=None2026.03 | 66 | |
| STTSModel=STTS, Pruning Strength=30%2026.03 | 63.4 | |
| BaselineModel=Baseline, Pruning Strength=0%2026.03 | 62.8 | |
| STTSModel=STTS, Pruning Strength=40%2026.03 | 62.4 | |
| STTSModel=STTS, Pruning Strength=50%2026.03 | 62.4 | |
| PLM-8BModel=PLM-8B, Pruning Strength=None2026.03 | 58.3 |