Sunflower: A New Approach To Expanding Coverage of African Languages in Large Language Models

2025-10-08

AI Summary

African languages remain systematically marginalized in mainstream large language models (LLMs), particularly in highly linguistically diverse countries such as Uganda. Method: We propose a region-centric language support paradigm—departing from fragmented, language-by-language fine-tuning—to enable unified modeling and efficient coverage of the vast majority of indigenous Ugandan languages. Leveraging the Qwen3 architecture, we develop the open-source Sunflower-14B/32B bilingual-scale models, integrating multilingual pretraining, cross-lingual transfer leveraging linguistic similarity, and open-data augmentation. Results: Empirical evaluation demonstrates substantial improvements in understanding across low-resource Ugandan languages, achieving state-of-the-art performance. The approach significantly mitigates language barriers in critical applications and establishes a reusable, systematic methodology for AI adaptation to African languages.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Computer Vision: Language and Vision

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

There are more than 2000 living languages in Africa, most of which have been bypassed by advances in language technology. Current leading LLMs exhibit strong performance on a number of the most common languages (e.g. Swahili or Yoruba), but prioritise support for the languages with the most speakers first, resulting in piecemeal ability across disparate languages. We contend that a regionally focussed approach is more efficient, and present a case study for Uganda, a country with high linguistic diversity. We describe the development of Sunflower 14B and 32B, a pair of models based on Qwen 3 with state of the art comprehension in the majority of all Ugandan languages. These models are open source and can be used to reduce language barriers in a number of important practical applications.

Problem

• Expanding African language coverage in large language models
• Addressing piecemeal language support across disparate African languages
• Developing regionally focused models for high linguistic diversity countries

Innovation

• Regionally focused approach for African language coverage
• Models based on Qwen 3 architecture with enhancements
• Open source models supporting Ugandan language applications