TildeOpen LLM: Leveraging Curriculum Learning to Achieve Equitable Language Representation
This paper addresses the problem of linguistic inequity and underperformance of LLMs for speakers of low-resource European languages by providing a more balanced and effective multilingual model.
Large language models often underperform in European languages due to training data imbalance. TildeOpen LLM, a 30-billion-parameter model trained on 34 European languages, addresses this by combining dataset upsampling with a curriculum-based training schedule. The model achieves up to a tenfold reduction in linguistic errors compared to leading baselines, particularly for Baltic, Finno-Ugric, and Slavic languages.
Large language models often underperform in many European languages due to the dominance of English and a few high-resource languages in training data. This paper presents TildeOpen LLM, a 30-billion-parameter open-weight foundational model trained for 34 European languages to promote linguistic equity and improve performance for low-resource languages. To address the data imbalance, we combine dataset upsampling with a curriculum-based training schedule that alternates between uniform and natural language distributions. The resulting model performs favorably compared to other multilingual LLMs despite being trained with significantly fewer computing resources. Evaluation across multiple multilingual benchmarks shows that TildeOpen surpasses existing open-weight models in text generation and comprehension, particularly for Baltic, Finno-Ugric, and Slavic languages. Human evaluations confirm an up to tenfold reduction in linguistic errors relative to leading baselines. The model and associated resources are fully open-weight and publicly available at huggingface.co/TildeAI/TildeOpen-30b. These outcomes demonstrate that careful data curation and balanced training strategies can substantially enhance multilingual model quality without increasing model size or training volume.