TildeOpen LLM: Leveraging Curriculum Learning to Achieve Equitable Language Representation

TL;DR

TildeOpen LLM is a 30-billion-parameter multilingual model trained on 34 European languages, using curriculum learning and data balancing to improve performance in low-resource languages, and is openly available.

Contribution

The paper introduces a novel training strategy combining dataset upsampling and curriculum scheduling to enhance multilingual language model performance.

Findings

Outperforms existing open multilingual models in text generation and comprehension.

Achieves up to tenfold reduction in linguistic errors for low-resource languages.

Demonstrates effective multilingual performance with fewer training resources.

Abstract

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.