MameLoshnLM is the first open‑source 8B‑parameter language model dedicated to Yiddish, filling a gap left by multilingual corpora that often contain noisy Hans. The authors also introduce Oytser, a new evaluation benchmark designed to provide reliable, language‑specific metrics for Yiddish NLP. Together, the model and benchmark aim to accelerate development in a language with a rich textual tradition but limited digital resources.

Read original