Text-to-image retrieval on Winoground
0.133R@1 (T2I)IL-CLIP
Evaluation Results
| Method | Links | |
|---|---|---|
| IL-CLIPPre-training Dataset=CC3M2024.04 | 0.133 | |
| NegCLIPPre-training Dataset=CC3M2024.04 | 0.118 | |
| IL-CLIPPre-training Dataset=CC12M2024.04 | 0.101 | |
| Codebook-CLIPPre-training Dataset=CC12M2024.04 | 0.095 | |
| Codebook-CLIPPre-training Dataset=CC3M2024.04 | 0.088 | |
| CLIPPre-training Dataset=CC3M2024.04 | 0.081 | |
| NegCLIPPre-training Dataset=CC12M2024.04 | 0.073 | |
| CLIPPre-training Dataset=CC12M2024.04 | 0.072 |