Published: March 2024
We're excited to introduce spBLEU-1K, a new translation metric covering over 1,000 languages. It improves on traditional BLEU with SentencePiece-based scoring and broad support for low-resource languages.
Read More
Published: March 2024
Introducing AfroLingu-MT, a comprehensive new benchmark for African machine translation. It's designed to support inclusive, real-world MT research as part of our ACL 2024 Toucan release.
Explore Benchmark
Published: March 2024
Meet Cheetah – our new language model for Natural Language Generation in 517 African languages. Trained on a 42GB curated corpus and evaluated on AfroNLG, Cheetah sets a new standard for inclusive multilingual NLP.
Explore Cheetah
Published: March 2024
We’re proud to launch SERENGETI, a new suite of pretrained language models built for African NLP. Covering 517 languages, it sets a new benchmark for inclusion, performance, and linguistic diversity.
Discover SERENGETI