Image-Text Retrieval on DOCCI-FG sentence-level
27T2I Recall@1GoldiCLIP
Evaluation results
| Method | Links | ||
|---|---|---|---|
| GoldiCLIPData size=30M, Evaluation setting=Zero-shot2026.03 | 27 | 60.2 | |
| FLAIRData size=30M, Evaluation setting=Zero-shot2026.03 | 25 | 59 | |
| SigLIP 2Data size=10B, Evaluation setting=Zero-shot2026.03 | 23.2 | 61.2 | |
| COSMOSData size=30M, Evaluation setting=Zero-shot2026.03 | 23.1 | 57.3 | |
| PEcoreData size=5B, Evaluation setting=Zero-shot2026.03 | 22.4 | 61.8 | |
| DreamLIPData size=30M, Evaluation setting=Zero-shot2026.03 | 21.6 | 51.2 | |
| SigLIPData size=10B, Evaluation setting=Zero-shot2026.03 | 20.6 | 57.5 | |
| GoldiCLIPData size=3M, Evaluation setting=Zero-shot2026.03 | 18.7 | 43 | |
| OpenCLIPData size=2B, Evaluation setting=Zero-shot2026.03 | 17.4 | 49.7 | |
| FLAIRData size=3M, Evaluation setting=Zero-shot2026.03 | 15.2 | 35.7 | |
| COSMOSData size=3M, Evaluation setting=Zero-shot2026.03 | 14.2 | 32.9 | |
| DreamLIPData size=3M, Evaluation setting=Zero-shot2026.03 | 10.3 | 23.3 |