Audiovisual Dialogue Description on DiaDemBench
97.9REFHuman (avg. of 3 authors)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Human (avg. of 3 authors)2026.01 | 97.9 | 97.1 | |
| DiaDemSize=7B2026.01 | 65.9 | 79.3 | |
| Gemini-2.5-Pro2026.01 | 63.6 | 74.8 | |
| Gemini-3-Pro2026.01 | 63.1 | 71 | |
| Gemini-2.5-Flash2026.01 | 58.5 | 73.3 | |
| Qwen3-Omni-CaptionerSize=30B-A3B2026.01 | 43.9 | 58.8 | |
| AVoCaDOSize=7B2026.01 | 38.7 | 51.7 | |
| Qwen3-Omni-InstructSize=30B-A3B2026.01 | 36.8 | 47.5 | |
| ARC-Qwen-Video-NarratorSize=7B2026.01 | 32.8 | 48.5 | |
| UGC-VideoCaptionerSize=3B2026.01 | 29.7 | 47 | |
| Qwen2.5-OmniSize=7B2026.01 | 26.1 | 37.1 | |
| OmniVinciSize=9B2026.01 | 17.1 | 25.2 | |
| HumanOmniV2Size=7B2026.01 | 15.3 | 21.5 | |
| ARC-Qwen-VideoSize=7B2026.01 | 11.7 | 17.2 | |
| video-SALMONN-2Size=7B2026.01 | 11.5 | 16.6 |