Pairwise Preference Prediction on DyCoBench-1K Multiple Criteria
65.6Preference Rate (A > B)DYCORM(OURS)
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| DYCORM(OURS)Model Category=Reward Models2026.05 | 65.6 | 67.8 | 57.8 | 66.2 | |
| GPT-5 (25-08-07)Model Category=General LMMs2026.05 | 57.1 | 58.4 | 54.2 | 57.6 | |
| GEMINI-3.0-PROModel Category=General LMMs2026.05 | 56.7 | 58 | 52.8 | 57.2 | |
| GPT-4O (24-11-20)Model Category=General LMMs2026.05 | 56 | 57.3 | 51.1 | 55.5 | |
| INTERNVL3.5-14BModel Category=General LMMs2026.05 | 52.6 | 53.9 | 47.7 | 52.1 | |
| QWEN3VL-8BModel Category=General LMMs2026.05 | 51.9 | 53.2 | 47 | 51.4 | |
| INTERNVL3.5-8BModel Category=General LMMs2026.05 | 51.2 | 52.5 | 46.3 | 50.7 | |
| LLAVA-ONEVISION-1.5-7BModel Category=General LMMs2026.05 | 51 | 52.3 | 46.1 | 50.5 | |
| QWEN3VL-4BModel Category=General LMMs2026.05 | 50.7 | 52 | 45.8 | 50.2 | |
| HPSV3Model Category=Reward Models2026.05 | 49.3 | 50.6 | 44.4 | 48.8 | |
| HPSV2Model Category=Reward Models2026.05 | 49 | 50.3 | 44.1 | 48.5 | |
| MPSModel Category=Reward Models2026.05 | 48.7 | 50 | 43.8 | 48.2 | |
| IMAGEREWARDModel Category=Reward Models2026.05 | 48.5 | 49.8 | 43.6 | 48 | |
| PICKSCOREModel Category=Reward Models2026.05 | 47.2 | 48.5 | 42.3 | 46.7 | |
| HPSModel Category=Reward Models2026.05 | 47 | 48.3 | 42.1 | 46.5 | |
| AESTHETIC SCORE PREDICTORModel Category=Reward Models2026.05 | 46.6 | 47.9 | 41.7 | 46.1 | |
| CLIPSCOREModel Category=Reward Models2026.05 | 46.3 | 47.6 | 41.4 | 45.8 |