Long-context Video Understanding on MLVU MCQ (val)
70.3ScoreBaseline
Evaluation results
| Method | Links | |
|---|---|---|
| BaselineModel=Baseline, Pruning Strength=0%2026.03 | 70.3 | |
| STTSModel=STTS, Pruning Strength=30%2026.03 | 69.5 | |
| STTSModel=STTS, Pruning Strength=50%2026.03 | 68.4 | |
| STTSModel=STTS, Pruning Strength=40%2026.03 | 67.5 | |
| Qwen3-VL-4BModel=Qwen3-VL-4B, Pruning Strength=None2026.03 | 58.4 | |
| InternVL3.5-8BModel=InternVL3.5-8B, Pruning Strength=None2026.03 | 53.2 | |
| PLM-8BModel=PLM-8B, Pruning Strength=None2026.03 | 52.6 |