DECL-SGRAF

Reported on 42 benchmarks across 3 tasks

Note: results are matched by exact model name. Different papers may use the same name for different model variants.

Miscellaneous28 results

Image Retrieval with Multi-Modal QueryonCC152K
Image-to-text R@1
39
best: 43.6 (UGNCL)
Image Retrieval with Multi-Modal QueryonCC152K
Image-to-text R@10
75.5
best: 78.1 (ReCon)
Image Retrieval with Multi-Modal QueryonCC152K
Image-to-text R@5
66.1
best: 68.7 (ReCon)
Image Retrieval with Multi-Modal QueryonCC152K
R-Sum
364.3
best: 380.5 (ReCon)
Image Retrieval with Multi-Modal QueryonCC152K
Text-to-image R@1
40.7
best: 44.9 (ReCon)
Image Retrieval with Multi-Modal QueryonCC152K
Text-to-image R@10
76.7
best: 78.4 (CRCL)
Image Retrieval with Multi-Modal QueryonCC152K
Text-to-image R@5
66.3
best: 68.4 (UGNCL)
Image Retrieval with Multi-Modal QueryonFlickr30K-Noisy
Image-to-text R@1
77.5
best: 80.3 (ReCon)
Image Retrieval with Multi-Modal QueryonFlickr30K-Noisy
Image-to-text R@10
97
best: 98.3 (CTPR-SGR)
Image Retrieval with Multi-Modal QueryonFlickr30K-Noisy
Image-to-text R@5
93.8
best: 95.8 (CTPR-SGR)
Image Retrieval with Multi-Modal QueryonFlickr30K-Noisy
R-Sum
494.7
best: 511.8 (ReCon)
Image Retrieval with Multi-Modal QueryonFlickr30K-Noisy
Text-to-image R@1
56.1
best: 61.6 (ReCon)
Image Retrieval with Multi-Modal QueryonFlickr30K-Noisy
Text-to-image R@10
88.5
best: 92.7 (CTPR-SGR)
Image Retrieval with Multi-Modal QueryonFlickr30K-Noisy
Text-to-image R@5
81.8
best: 85.5 (ReCon)
Cross-Modal Information RetrievalonCC152K
Image-to-text R@1
39
best: 43.6 (UGNCL)
Cross-Modal Information RetrievalonCC152K
Image-to-text R@10
75.5
best: 78.1 (ReCon)
Cross-Modal Information RetrievalonCC152K
Image-to-text R@5
66.1
best: 68.7 (ReCon)
Cross-Modal Information RetrievalonCC152K
R-Sum
364.3
best: 380.5 (ReCon)
Cross-Modal Information RetrievalonCC152K
Text-to-image R@1
40.7
best: 44.9 (ReCon)
Cross-Modal Information RetrievalonCC152K
Text-to-image R@10
76.7
best: 78.4 (CRCL)
Cross-Modal Information RetrievalonCC152K
Text-to-image R@5
66.3
best: 68.4 (UGNCL)
Cross-Modal Information RetrievalonFlickr30K-Noisy
Image-to-text R@1
77.5
best: 80.3 (ReCon)
Cross-Modal Information RetrievalonFlickr30K-Noisy
Image-to-text R@10
97
best: 98.3 (CTPR-SGR)
Cross-Modal Information RetrievalonFlickr30K-Noisy
Image-to-text R@5
93.8
best: 95.8 (CTPR-SGR)
Cross-Modal Information RetrievalonFlickr30K-Noisy
R-Sum
494.7
best: 511.8 (ReCon)
Cross-Modal Information RetrievalonFlickr30K-Noisy
Text-to-image R@1
56.1
best: 61.6 (ReCon)
Cross-Modal Information RetrievalonFlickr30K-Noisy
Text-to-image R@10
88.5
best: 92.7 (CTPR-SGR)
Cross-Modal Information RetrievalonFlickr30K-Noisy
Text-to-image R@5
81.8
best: 85.5 (ReCon)

Natural Language Processing14 results

Cross-Modal RetrievalonCC152K
Image-to-text R@1
39
best: 43.6 (UGNCL)
Cross-Modal RetrievalonCC152K
Image-to-text R@10
75.5
best: 78.1 (ReCon)
Cross-Modal RetrievalonCC152K
Image-to-text R@5
66.1
best: 68.7 (ReCon)
Cross-Modal RetrievalonCC152K
R-Sum
364.3
best: 380.5 (ReCon)
Cross-Modal RetrievalonCC152K
Text-to-image R@1
40.7
best: 44.9 (ReCon)
Cross-Modal RetrievalonCC152K
Text-to-image R@10
76.7
best: 78.4 (CRCL)
Cross-Modal RetrievalonCC152K
Text-to-image R@5
66.3
best: 68.4 (UGNCL)
Cross-Modal RetrievalonFlickr30K-Noisy
Image-to-text R@1
77.5
best: 80.3 (ReCon)
Cross-Modal RetrievalonFlickr30K-Noisy
Image-to-text R@10
97
best: 98.3 (CTPR-SGR)
Cross-Modal RetrievalonFlickr30K-Noisy
Image-to-text R@5
93.8
best: 95.8 (CTPR-SGR)
Cross-Modal RetrievalonFlickr30K-Noisy
R-Sum
494.7
best: 511.8 (ReCon)
Cross-Modal RetrievalonFlickr30K-Noisy
Text-to-image R@1
56.1
best: 61.6 (ReCon)
Cross-Modal RetrievalonFlickr30K-Noisy
Text-to-image R@10
88.5
best: 92.7 (CTPR-SGR)
Cross-Modal RetrievalonFlickr30K-Noisy
Text-to-image R@5
81.8
best: 85.5 (ReCon)