Audio-Visual Segment Narration on EgoAVU-Bench
2.63S ScoreEgoAVU-Instruct (Full)
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| EgoAVU-Instruct (Full)Size=7B, Tuning=Full2026.02 | 2.63 | 22.68 | 28.7 | |
| EgoAVU-Instruct (LoRA)Size=7B, Tuning=LoRA2026.02 | 2.45 | 22.53 | 28.34 | |
| MiniCPM-oSize=8B2026.02 | 2.06 | 9.68 | 12.19 | |
| Qwen2.5-OmniSize=7B2026.02 | 1.99 | 9.99 | 13.39 | |
| Qwen2.5-OmniSize=3B2026.02 | 1.85 | 8.63 | 13.08 | |
| Intern-OmniSize=8.7B2026.02 | 1.82 | 8.69 | 14.7 | |
| Baichuan-OmniSize=7B2026.02 | 1.79 | 8.34 | 14.21 | |
| VideoLLaMA2Size=7B2026.02 | 1.71 | 7.5 | 13.89 | |
| Phi4-mmSize=8B2026.02 | 1.69 | 12.17 | 16.9 |