Image-Text Retrieval on Winoground (test)
74Text ScoreYUKINO
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| YUKINOPre-training Dataset=LAION2B2024.12 | 74 | 46.25 | 42.75 | |
| YUKINO-OTIPre-training Dataset=LAION2B2024.12 | 66.75 | 42.5 | 37.5 | |
| GPT-4V+CoCoT2024.12 | 58.5 | 49.5 | 44.5 | |
| GPT-4V2024.12 | 54.5 | 42.5 | 37.75 | |
| LLaVA-1.52024.12 | 36 | 33.25 | 20.05 | |
| CLIPPre-training Dataset=LAION2B2024.12 | 34.75 | 11 | 7.5 | |
| TripletCLIPPre-training Dataset=CC12M2024.12 | 23.25 | 6.25 | 4.25 | |
| Structure-CLIPPre-training Dataset=LAION2B2024.12 | 22.5 | 14 | 8.25 | |
| Neg-CLIPPre-training Dataset=LAION2B2024.12 | 20.5 | 12.25 | 6.5 | |
| Neg-CLIPPre-training Dataset=CC12M2024.12 | 18.25 | 6.5 | 4.25 |