Video Understanding on Vinoground video sub-task zero-shot
38.2Accuracy (zero-shot)GPT-4o
Evaluation results
| Method | Links | |
|---|---|---|
| GPT-4o2025.07 | 38.2 | |
| InternVL3.5LLM=Qwen3-14B, Frames=16, Tokens=3.96B2025.07 | 31.2 | |
| InternVL3.5-REGATELLM=Qwen3-14B, Frames=16, Tokens=2.32B (↓ 41.41%)2025.07 | 31.2 | |
| LLaVA-OneVisionLLM=Qwen2-7B, Frames=322025.07 | 29.4 | |
| Video-LLaVALLM=Vicuna-7B, Frames=82025.07 | 25.8 | |
| LLaVA-NeXT-VideoLLM=Vicuna-7B, Frames=322025.07 | 25.6 | |
| VideoLLaMA2-REGATELLM=Qwen2-7B, Frames=16, Tokens=49.27M (↓ 41.22%)2025.07 | 25.2 | |
| VideoLLaMA2LLM=Qwen2-7B, Frames=16, Tokens=83.82M2025.07 | 24.6 | |
| VideoChat2-REGATELLM=Mistral-7B, Frames=16, Tokens=2.22B (↓ 43.51%)2025.07 | 22.8 | |
| Gemini-1.5-Pro2025.07 | 22.6 | |
| VideoChat2LLM=Mistral-7B, Frames=16, Tokens=3.93B2025.07 | 22 |