NbAiLab/nb-notram-llama-3.2-3b-instruct
NbAiLab/nb-notram-llama-3.2-3b-instruct is a 3.2 billion parameter instruction-tuned causal language model developed by the National Library of Norway (NB-AiLab). Built upon Meta's Llama-3.2-3B-Instruct, it is specifically fine-tuned to enhance instruction-following in Norwegian Bokmål and Nynorsk while maintaining strong English performance. This model is notable for being trained exclusively on publicly available data, making it suitable for dialogue systems, summarization, and Q&A in its target languages with a tendency for concise answers.
Loading preview...
NbAiLab/nb-notram-llama-3.2-3b-instruct Overview
"nb-notram-llama-3.2-3b-instruct" is a 3.2 billion parameter model from the National Library of Norway (NB-AiLab), part of their "NB-Llama-3.x" and "NoTraM" series. It is built on Meta's "Llama-3.2-3B-Instruct" and has been fine-tuned to significantly improve instruction-following in Norwegian Bokmål and Norwegian Nynorsk, while also preserving its strong English capabilities. A key differentiator is its exclusive training on publicly available data, explicitly excluding legal deposit material, making it a transparent and reproducible resource for Norwegian language adaptation.
Key Capabilities
- Multilingual Instruction Following: Excels in understanding and executing instructions in Norwegian Bokmål, Norwegian Nynorsk, and English.
- Concise Responses: The model is tuned to produce shorter, more direct answers, which can be beneficial for specific application types.
- Public Data Training: Developed entirely using publicly accessible datasets, ensuring transparency and reproducibility.
- Robust Base: Leverages the strong instruction-following foundation of "Llama-3.2-3B-Instruct" with a light preference optimization step.
Good For
- Norwegian Dialogue Systems: Ideal for creating assistant-style applications and chatbots in Norwegian Bokmål and Nynorsk.
- Summarization and Q&A: Effective for generating summaries and answering questions in both Norwegian dialects.
- Research on Small Language Adaptation: Useful for exploring techniques to adapt instruction-tuned models to smaller languages using public data, aiming to reduce "knowledge pocketing".