Text-vision Alignment on Q-Eval-Image 100K (test)
0.822Instance SRCCQ-Eval-Score
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Q-Eval-Scorebackbone=Qwen2-VL-7B-Instruct, zero-shot=false2025.03 | 0.822 | 0.802 | 0.964 | 0.969 | |
| CLIPScorezero-shot=false2025.03 | 0.768 | 0.74 | 0.958 | 0.956 | |
| BLIP2Scorezero-shot=false2025.03 | 0.766 | 0.743 | 0.933 | 0.934 | |
| ImageRewardzero-shot=false2025.03 | 0.762 | 0.732 | 0.925 | 0.955 | |
| VQAScorezero-shot=true, official default weight=true2025.03 | 0.549 | 0.468 | 0.385 | 0.555 | |
| BLIP2Scorezero-shot=true, official default weight=true2025.03 | 0.297 | 0.33 | 0.764 | 0.835 | |
| CLIPScorezero-shot=true, official default weight=true2025.03 | 0.245 | 0.252 | 0.617 | 0.685 |