Video Question Answering on OmniVideo (test)
100Overall ScoreHuman
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Human2026.06 | 100 | — | — | — | — | — | |
| Gemini-3.1-ProSampling frames=642026.06 | 83.96 | 83.62 | 84.5 | 83.21 | 82.61 | 84.59 | |
| Ours(Qwen3-Omni)Size=30B, Sampling frames=64, Finetuned on=OmniVideo-100K2026.06 | 63.56 | 60.34 | 67.05 | 59.54 | 62.11 | 64.24 | |
| Ours(VITA-1.5)Size=8B, Sampling frames=16, Finetuned on=OmniVideo-100K2026.06 | 61.58 | 59.48 | 63.18 | 60.31 | 59.01 | 62.79 | |
| Ours(Qwen2.5-Omni)Size=7B, Sampling frames=64, Finetuned on=OmniVideo-100K2026.06 | 60.59 | 62.93 | 62.4 | 54.96 | 54.66 | 63.37 | |
| MiniCPM-o 4.5Size=9B, Sampling frames=642026.06 | 55.25 | 56.9 | 58.14 | 48.09 | 53.42 | 56.1 | |
| Qwen3-OmniSize=30B, Sampling frames=642026.06 | 49.7 | 43.1 | 55.04 | 45.04 | 49.07 | 50 | |
| OmniVinciSize=9B, Sampling frames=642026.06 | 47.13 | 43.97 | 51.55 | 41.22 | 47.2 | 47.09 | |
| Uni-MoE-2.0-OmniSize=30B, Sampling frames=642026.06 | 46.93 | 39.66 | 52.71 | 41.98 | 46.58 | 47.09 | |
| video-SALMONN 2+Size=7B, Sampling frames=642026.06 | 45.15 | 37.93 | 49.22 | 43.51 | 46.58 | 44.48 | |
| Qwen2.5-OmniSize=7B, Sampling frames=642026.06 | 42.77 | 39.66 | 46.51 | 38.17 | 38.51 | 44.77 | |
| VITA-1.5Size=8B, Sampling frames=162026.06 | 40.99 | 28.45 | 43.02 | 48.09 | 44.72 | 39.24 |