Human Preference Alignment on REACT-Video
61Acc (Tie, Overall)REACT
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| REACTModel Category=Our REACT2026.01 | 61 | — | — | — | — | 81.3 | |
| UnifiedRewardModel Category=Video Evaluator2026.01 | 41.6 | 39 | 40 | 70.7 | 67.4 | 70.1 | |
| VideoRewardModel Category=Video Evaluator2026.01 | 41.5 | 40.7 | 41.7 | 52.4 | 57.2 | 55.1 | |
| Gemini-2.5-FlashModel Category=General Multimodal Language Model, Sampling Rate=2 fps2026.01 | 38.4 | — | — | — | — | 55.3 | |
| Q-insightModel Category=Image Evaluator2026.01 | 38.4 | — | — | — | — | 55.9 | |
| Qwen2.5-VL-7BModel Category=General Multimodal Language Model, Sampling Rate=2 fps2026.01 | 37.6 | — | — | — | — | 50.9 | |
| VisualQuality-R1Model Category=Image Evaluator2026.01 | 37.6 | — | — | — | — | 61 | |
| VisualQuality-R1Model Category=Image Evaluator, Prompting Strategy=Refined prompt (+Rep)2026.01 | 37.6 | — | — | — | — | 58.6 | |
| Gemini-2.5-ProModel Category=General Multimodal Language Model, Sampling Rate=2 fps2026.01 | 37 | — | — | — | — | 53.4 | |
| Qwen2.5-VL-32BModel Category=General Multimodal Language Model, Sampling Rate=2 fps2026.01 | 36.4 | — | — | — | — | 58.3 | |
| Q-insightModel Category=Image Evaluator, Prompting Strategy=Refined prompt (+Rep)2026.01 | 35.4 | — | — | — | — | 55.2 | |
| VideoScore2Model Category=Video Evaluator2026.01 | 34.2 | 36.2 | 36.4 | 55 | 54 | 52.1 |