Image-Text Retrieval on MSCOCO (val)
55.5T2I Recall@1GoldiCLIP
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| GoldiCLIPData size=30M, Evaluation setting=Zero-shot2026.03 | 55.5 | 78.9 | 70.3 | 89.3 | |
| FLAIRData size=30M, Evaluation setting=Zero-shot2026.03 | 53.3 | 77.5 | 68 | 87.8 | |
| COSMOSData size=30M, Evaluation setting=Zero-shot2026.03 | 52.5 | 77.2 | 68 | 87.8 | |
| SigLIP 2Data size=10B, Evaluation setting=Zero-shot2026.03 | 52.5 | 76.6 | 70 | 88.3 | |
| PEcoreData size=5B, Evaluation setting=Zero-shot2026.03 | 49.9 | 74.8 | 70.3 | 88.9 | |
| SigLIPData size=30M, Evaluation setting=Zero-shot2026.03 | 49 | 74.5 | 64.1 | 86.2 | |
| SigLIPData size=10B, Evaluation setting=Zero-shot2026.03 | 47.2 | 72.1 | 65.5 | 86.2 | |
| GoldiCLIPData size=3M, Evaluation setting=Zero-shot2026.03 | 45.1 | 71.2 | 58.6 | 82.4 | |
| DreamLIPData size=30M, Evaluation setting=Zero-shot2026.03 | 44.8 | 69.8 | 62.3 | 84.5 | |
| OpenCLIPData size=2B, Evaluation setting=Zero-shot2026.03 | 41.7 | 67.1 | 59.3 | 82.4 | |
| Long-CLIPData size=400M, Evaluation setting=Zero-shot2026.03 | 40.3 | 65.9 | 57.3 | 80.8 | |
| COSMOSData size=3M, Evaluation setting=Zero-shot2026.03 | 40.1 | 66.8 | 53.1 | 78.3 | |
| LoTLIPData size=100M, Evaluation setting=Zero-shot2026.03 | 38.1 | 63.8 | 59.7 | 81.5 | |
| FLAIRData size=3M, Evaluation setting=Zero-shot2026.03 | 37.7 | 64.5 | 51.6 | 77.2 | |
| LiTData size=100M, Evaluation setting=Zero-shot2026.03 | 32.7 | 57.8 | 51.9 | 75.7 | |
| DreamLIPData size=3M, Evaluation setting=Zero-shot2026.03 | 29.8 | 55.4 | 40.8 | 68.4 |