Researchers release Yiddish-focused MameLoshnLM
The open-source 8B model uses a new corpus and benchmark for low-resource Yiddish NLP.
Why it matters
The release shows how targeted data curation and evaluation can improve modeling for low-resource languages that are poorly served by noisy multilingual web corpora. It also provides reusable resources for Yiddish NLP research and applications.
The key points
- 1.MameLoshnLM is an open-source 8B model for Yiddish.
- 2.Oytser combines web-native and literary Yiddish sources.
- 3.Kashes tests translation, analysis, extraction and understanding.
Researchers introduced MameLoshnLM, an open-source 8B-parameter language model built specifically for Yiddish by continuing pretraining on Llama 3.1 8B. The work also introduces Oytser, a higher-quality Yiddish pretraining corpus, and Kashes, a multi-task benchmark covering translation, linguistic analysis, information extraction and language understanding. The authors report that MameLoshnLM outperforms open baselines of similar scale across the benchmark.
⚡ Try this today
Use Kashes to evaluate Yiddish-language systems before relying on general multilingual models.
Sources & original reporting
This brief summarizes and links to reporting from the publishers below.
Enjoyed this brief? Get the next one in your inbox.
More in Research
New papers target VLMs' spatial reasoning gap
Researchers propose runtime memory, RL training, benchmarks and 3D generation methods for spatial AI.
New papers probe spatial intelligence in AI vision models
SpaRRTa, SMA and PinpointQA target spatial reasoning gaps in visual and multimodal systems.
Papers refine pseudo-labeling for semi-supervised vision
New arXiv papers target label scarcity in HD mapping and semantic segmentation.