List your company

ResearchFormerly WizWand

Machine Translation

Benchmarks

DatasetSOTA methodMetricTrendResultsLast updated

FLores De→En 0-shot

ADAFUSE (Top-2 Base)

45.25spBLEU

10

Feb 26, 2026

FLores En→De 0-shot

ADAFUSE (Top-2 Base)

39.83spBLEU

10

Feb 26, 2026

58.84chrF++ (English, Fwd)

10

Feb 26, 2026

41.49spBLEU (En ->)

10

Feb 26, 2026

50.75chrF

10

Feb 26, 2026

human-collected unfaithful translation En ⇒ Zh (test)

Target-constrained tuning LoRA

19.1BLEU

10

Feb 26, 2026

human-collected unfaithful translation Zh ⇒ En (test)

R-Drop tuning LORA

16.6BLEU

10

Feb 26, 2026

human-collected unfaithful translation En ⇒ De (test)

Target-constrained tuning

20.9BLEU

10

Feb 26, 2026

human-collected unfaithful translation De ⇒ En (test)

Target-constrained tuning LoRA

30.8BLEU

10

Feb 26, 2026

WMT en→xx (test)

32.17BLEU

10

Feb 26, 2026

WMT xx-en Average

37.36BLEU

10

Feb 26, 2026

English-Nigerian-Pidgin MT Data

63.9CHRF

10

Feb 26, 2026

English-Luganda MT Data

49.38CHRF

10

Feb 26, 2026

0.775COMET Score

10

Feb 26, 2026

0.677COMET

10

Feb 26, 2026

Ambiguous Constraint En-Zh (test)

Stage1 + Stage2 Tem.

31.4SacreBLEU

10

Feb 26, 2026

Ambiguous Constraint De-En (test)

Stage1 + Stage2 Tem.

35.1SacreBLEU

10

Feb 26, 2026

Ambiguous Constraint En-Zh 1.0 (test)

Stage1 + Stage2 Tem.

31.4SacreBLEU

10

Feb 26, 2026

Ambiguous Constraint De-En 1.0 (test)

Stage1 + Stage2 Tem.

35.1SacreBLEU

10

Feb 26, 2026

31.9BLEU Score

10

Feb 26, 2026

32.9BLEU

10

Feb 26, 2026

Out-Of-Domain (OOD) (val)

11.08BLEU

10

Feb 27, 2026

Adapted-Domain (AD) (val)

Seq2Seq_ID + Seq2Seq_AD

23.35BLEU

10

Feb 27, 2026

In-Domain (ID) (val)

Seq2Seq_ID + Seq2Seq_AD

40.72BLEU

10

Feb 27, 2026

IT En-De out-of-domain WMT14 (test)

27.5BLEU

10

Feb 26, 2026