Vision-Language Understanding on Winoground
61.5Text AccuracyCROCScore
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| CROCScorefine-tuned=true2025.05 | 61.5 | 55.8 | 46.8 | |
| VQAScore2025.05 | 60.5 | 57.5 | 45.8 | |
| CROCScoreplainfine-tuned=false2025.05 | 58.7 | 52.2 | 42.8 | |
| BLIP2-ITM2025.05 | 42.8 | 22.3 | 18.3 | |
| AlignScore2025.05 | 35.3 | 10.5 | 8.25 |