Video understanding and ordering on Advertisement Video Editing Dataset
66.48CSAQwen2.5-VL–32B
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Qwen2.5-VL–32BStrategy=Instruct2026.03 | 66.48 | 2.472 | 0.998 | |
| AutoCut2026.03 | 65.93 | 10.714 | 1.036 | |
| Qwen3–8BStrategy=Caption + SFT2026.03 | 56.87 | 3.022 | 1.1227 | |
| GPT-4o + MGSV2026.03 | 26.89 | 7.756 | 1.1364 | |
| Qwen2.5-VL–7BStrategy=Instruct2026.03 | 24.18 | 1.648 | 0.9649 | |
| Qwen3–32BStrategy=Caption2026.03 | 17.31 | 1.373 | 0.9342 | |
| Qwen3–8BStrategy=Caption2026.03 | 13.74 | 1.648 | 0.9308 | |
| InternVL-3.5-8B2026.03 | 2.47 | 1.653 | 0.9466 | |
| LLaVA-v1.6-Mistral-7B-HF2026.03 | 0.27 | 0.923 | 0.9914 |