Mistral-Nemo-Instruct-Uz
| Property | Value |
|---|---|
| License | Apache 2.0 |
| Languages | Uzbek, English |
| Base Model | Mistral-Nemo-Instruct-2407 |
What is Mistral-Nemo-Instruct-Uz?
Mistral-Nemo-Instruct-Uz is a specialized language model designed to bridge the gap between Uzbek and English language processing. Built upon the Mistral-Nemo-Instruct-2407 architecture, this model has been extensively pre-trained and instruction-tuned using a combination of public and synthetic datasets in both Uzbek and English languages.
Implementation Details
The model demonstrates impressive performance metrics, particularly in translation tasks. It achieves a BLEU score of 30.49 for Uzbek-to-English translation and 15.52 for English-to-Uzbek translation, representing significant improvements over the base model. The COMET scores of 87.04 (Uz-En) and 88.01 (En-Uz) further validate its translation capabilities.
- Utilizes 6 distinct datasets including tahrirchi/uz-crawl and Wikipedia-uzbek
- Implements text-generation-inference pipeline
- Maintains robust performance across multiple NLP tasks
Core Capabilities
- Bilingual translation between Uzbek and English
- Text summarization in both languages
- Question-answering capabilities
- Sentiment analysis (82.05% accuracy on Uzbek texts)
- News classification (58.2% accuracy on Uzbek news)
Frequently Asked Questions
Q: What makes this model unique?
The model stands out for its specialized optimization for the Uzbek language while maintaining strong English language capabilities. It shows significant improvements over base models in translation tasks while preserving performance on general language understanding benchmarks like MMLU.
Q: What are the recommended use cases?
The model excels in Uzbek-English translation tasks, content summarization, and dialogue systems. It's particularly useful for applications requiring natural language understanding in both Uzbek and English contexts, such as content localization, automated translation services, and multilingual chatbots.
