Video Preference Modeling on GenAI-Bench (evaluation)
68.7TauVR-THINKER
Evaluation results
| Method | Links | ||
|---|---|---|---|
| VR-THINKERSize=7B, Protocol=Reasoning-based Reward Models2025.10 | 68.7 | 82.3 | |
| UnifiedReward-ThinkSize=7B, Protocol=Reasoning-based Reward Models2025.10 | 64.7 | 80.4 | |
| UnifiedRewardSize=7B, Protocol=Generative-based Reward Models2025.10 | 61.2 | 76.8 | |
| VisionRewardSize=13B, Protocol=Classifier-based Reward Models2025.10 | 52.6 | 72.7 | |
| VideoRewardSize=2B, Protocol=Classifier-based Reward Models2025.10 | 49.9 | 73.1 | |
| VideoScoreSize=7B, Protocol=Classifier-based Reward Models2025.10 | 47.5 | 70.9 | |
| LiFTSize=13B, Protocol=Generative-based Reward Models2025.10 | 38.1 | 59.4 |