Video Question Answering on MSR-VTT (Frame Sampling Accuracy)
45.2Accuracy (50% Frames)SLAP
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| SLAPzero-shot=true2026.05 | 45.2 | 43.9 | 41.8 | -3.4 | |
| Video-LLaMA 3zero-shot=true2026.05 | 44.5 | 41.2 | 34.7 | -9.8 | |
| Stable Video Diffusionzero-shot=true2026.05 | 43.8 | 39.5 | 35.2 | -8.6 | |
| Linear Interpolationzero-shot=true2026.05 | 40.1 | 35.8 | 30.1 | -10 | |
| Zero-Order Holdzero-shot=true2026.05 | 38.4 | 31.2 | 22.5 | -15.9 |