UGNCL: Uncertainty-Guided Noisy Correspondence Learning for Efficient Cross-Modal Matching

Benchmark Model Rank Results
cross-modal-retrieval-with-noisy-1UGNCL#5R-Sum: 373.1Image-to-text R@1: 43.6Image-to-text R@5: 67.1
cross-modal-retrieval-with-noisy-2UGNCL#4R-Sum: 505.6Image-to-text R@1: 78.4Image-to-text R@5: 95.8
cross-modal-retrieval-with-noisy-3UGNCL#2R-Sum: 526.3Image-to-text R@1: 79.5Image-to-text R@5: 97.2