Does the Performance of Text-to-Image Retrieval Models Generalize Beyond Captions-as-a-Query?

Benchmark Model Rank Results
image-retrieval-on-conqa-conceptualCLIP#1Recall@1: 12.2Recall@5: 30.6Recall@10: 36.7R-precision: 6.8
image-retrieval-on-conqa-conceptualBLIP 2#2Recall@1: 8.2Recall@5: 28.6Recall@10: 36.7R-precision: 5.4
image-retrieval-on-conqa-conceptualBLIP#3Recall@1: 4.1Recall@5: 28.6Recall@10: 40.8R-precision: 5.4
image-retrieval-on-conqa-conceptualNAAF#4Recall@1: 4.1Recall@5: 12.2Recall@10: 16.3R-precision: 2.4
image-retrieval-on-conqa-conceptualSGRAF#5Recall@1: 0.0Recall@5: 8.2Recall@10: 10.2R-precision: 1.3
image-retrieval-on-conqa-descriptiveCLIP#1Recall@1: 20.7Recall@5: 58.3Recall@10: 65.5R-precision: 16.5
image-retrieval-on-conqa-descriptiveBLIP#2Recall@1: 20.7Recall@5: 58.3Recall@10: 62.1R-precision: 15.3
image-retrieval-on-conqa-descriptiveBLIP-2#3Recall@1: 20.7Recall@5: 51.7Recall@10: 62.1R-precision: 15.3
image-retrieval-on-conqa-descriptiveNAAF#4Recall@1: 13.8Recall@5: 34.5Recall@10: 44.8R-precision: 10.6
image-retrieval-on-conqa-descriptiveSGRAF#5Recall@1: 6.9Recall@5: 24.1Recall@10: 34.5R-precision: 7.9