NLLB-200
A multilingual machine translation model released by Meta in 2022, covering 200 languages with parameter scales ranging from 600M to 54B. It is designed to provide high-quality translation capabilities for low-resource languages.
Timeline (last 90 days)
The research team used COILD to conduct fine-tuning experiments on two multilingual neural machine translation models, IndicTrans2-Distilled and NLLB-200.
The research team benchmarked four machine translation systems across three architectural families: NLLB-200 (600M parameters), mT5-base, and two AfriNLLB variants.
Fine-tuning on MudawanSn yields significant improvements in both translation directions, with the AfriNLLB variants optimized for African languages outperforming the general-purpose NLLB-200 and mT5.
All Facts (3)
The research team used COILD to conduct fine-tuning experiments on two multilingual neural machine translation models, IndicTrans2-Distilled and NLLB-200.
50%UnverifiedThe research team benchmarked four machine translation systems across three architectural families: NLLB-200 (600M parameters), mT5-base, and two AfriNLLB variants.
50%UnverifiedFine-tuning on MudawanSn yields significant improvements in both translation directions, with the AfriNLLB variants optimized for African languages outperforming the general-purpose NLLB-200 and mT5.
50%