XCOMET-XXL
XCOMET-XXL is a next-generation neural translation quality evaluation metric based on the COMET framework, using a large multilingual model at its core to model semantic consistency and error spans between a translation and its reference. Compared with n-gram overlap metrics such as BLEU, XCOMET is more robust to word order changes and synonymous substitutions.
Timeline (last 90 days)
On Odia and Swahili, the back-translation method, using only 1000 parallel sentence pairs, outperforms continued pre-training by 7.7 to 16.4 XCOMET-XXL points in in-domain evaluation.
On all low-resource languages, back-translation, compared with fine-tuning on parallel data alone, improves the XCOMET-XXL score by 16% to 173% in the direction of translation into the low-resource language.
All Facts (2)
On Odia and Swahili, the back-translation method, using only 1000 parallel sentence pairs, outperforms continued pre-training by 7.7 to 16.4 XCOMET-XXL points in in-domain evaluation.
50%UnverifiedOn all low-resource languages, back-translation, compared with fine-tuning on parallel data alone, improves the XCOMET-XXL score by 16% to 173% in the direction of translation into the low-resource language.
50%