Long-form Video Understanding on LV-Bench
21.3ERMovieChat
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| MovieChatSize=7B, Evaluation Protocol=Zero-shot2025.12 | 21.3 | 23.1 | 25.9 | 22.3 | 24 | 17.2 | 22.5 | |
| VideoChatGPTSize=7B, Evaluation Protocol=Zero-shot2025.12 | 22.9 | 22.6 | 22.7 | 25.5 | 23.4 | 24.1 | 23.1 | |
| LLaVA-SFTSize=7B, Frames=16, Evaluation Protocol=Zero-shot2025.12 | 23.1 | 25.4 | 23.7 | 20.4 | 22.8 | 21.2 | 23.1 | |
| LWMSize=7B, Evaluation Protocol=Zero-shot2025.12 | 24.7 | 24.8 | 26.5 | 28.6 | 30.5 | 22.4 | 25.5 | |
| LLaVA-SFT + VideoScaffoldSize=7B, Frames=16, Evaluation Protocol=Zero-shot2025.12 | 24.7 | 26.6 | 25.3 | 26.1 | 29.7 | 27.7 | 24.5 | |
| LLaVA-OVSize=70B, Evaluation Protocol=Zero-shot2025.12 | 25 | 26.9 | 29.2 | 30.9 | 25.4 | 31 | 26.9 | |
| LLaMA-VIDSize=13B, Evaluation Protocol=Zero-shot2025.12 | 25.4 | 21.7 | 23.4 | 26.4 | 26.5 | 17.2 | 23.9 | |
| LLaVA-SFT + VideoScaffoldSize=7B, Frames=60, Evaluation Protocol=Zero-shot2025.12 | 29.2 | 28.1 | 28.5 | 28.2 | 33.3 | 29.3 | 29.1 | |
| LLaVA-NeXTSize=7B, Evaluation Protocol=Zero-shot2025.12 | 30.1 | 31.2 | 34.1 | 31.4 | 35 | 27.6 | 32.2 | |
| VideoScaffoldSize=7B, Frames=60, Evaluation Protocol=Zero-shot2025.12 | 32.1 | 31.2 | 29.6 | 30 | 35.3 | 25.9 | 31.5 |