Video Question Answering on EgoSchema ES-S (public subset)
55.8AccuracySF-VLM
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SF-VLMParam=13B, Video Frames=12024.03 | 55.8 | 1.89 | |
| SOTA (Zhang et al. 2023a)Param=20B, Video Frames=1802024.03 | 50.8 | 381 | |
| Just-LLMParam=7B, Video Frames=02024.03 | 45.8 | 0.41 | |
| Random2024.03 | 20 | — |