Reward Modeling on VideoScore-Bench v2
46Accuracy (Alignment)VideoScore2
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| VideoScore2Base Model=Qwen2.5-VL-7B2026.06 | 46 | 50 | 39.2 | 92.4 | 87.4 | 90.4 | 62.9 | 50.1 | 60.3 | |
| SG-PVRBase Model=Qwen3.5-9B2026.06 | 39.8 | 43.6 | 38.2 | 93.8 | 83.2 | 82.6 | 58.5 | 51.7 | 56.6 | |
| VideoScore2 (SFT Only)Base Model=Qwen2.5-VL-7B2026.06 | 39 | 41.4 | 35.6 | 87.8 | 81.4 | 83.8 | 48.2 | 42.4 | 53 | |
| VideoScore1.1Base Model=Mantis-8B2026.06 | 34.9 | 41.5 | 38.9 | 91 | 87 | 82.4 | 49 | 47 | 30.9 | |
| UnifiedReward2.0Base Model=Qwen3.5-9B2026.06 | 32.8 | — | 32.3 | — | 84.6 | 84 | — | 37.6 | 35.8 | |
| VideoRewardBase Model=Qwen2-VL-2B2026.06 | 29.6 | 23.8 | — | 61 | — | 67.4 | 41.4 | — | 44.6 | |
| OmniScoreBase Model=-2026.06 | 29.2 | — | — | — | — | 78.6 | — | — | 32.8 | |
| LiFTBase Model=VILA-13B2026.06 | 13.2 | 14.2 | — | 48.4 | — | 47.8 | 10.2 | — | 31.2 |