Multi30K
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
Multi30K MSCOCO
46.5BLEU
9
Multi30k En-De (MSCOCO)
56.64METEOR
9
Multi30K Task2 DE (test)
14.93FID
8
Multi30k En-De 2016 (test)
58.5METEOR
8
Multi30k En-Fr (MSCOCO)
70.51METEOR Score
8
Multi30k En-De
49.3BLEU@4
6
Multi30k Task1 (de-en)
38.38BLEU
6
Multi30k translated (test)
37.6BLEU-4
5
Multi30K SacreBLEU COMET (test)
40.6De SacreBLEU
4
Multi30K (test)
54.13BLEU
4
Multi30K En-De and De-En (test)
39.3Avg. BLEU
4
Multi30k (test)
19.38FPR@95
3
Multi30k zero-shot
80.9EN Mean Recall
3
Multi30k (val)
1.969PPL
2
Multi30k
1.43Test Loss (Mean)
2