Judging Quality Across Languages: A Multilingual Approach to Pretraining Data Filtering with Language Models
This addresses the bottleneck of multilingual data curation for researchers and practitioners training LLMs, though it appears incremental as an improved filtering method.
The paper tackles the problem of limited high-quality multilingual training data for pretraining large language models by introducing JQL, a systematic approach that uses lightweight annotators based on pretrained multilingual embeddings. The method outperforms heuristic filtering methods like Fineweb2 across 35 languages, enhancing downstream model training quality and increasing data retention rates.
High-quality multilingual training data is essential for effectively pretraining large language models (LLMs). Yet, the availability of suitable open-source multilingual datasets remains limited. Existing state-of-the-art datasets mostly rely on heuristic filtering methods, restricting both their cross-lingual transferability and scalability. Here, we introduce JQL, a systematic approach that efficiently curates diverse and high-quality multilingual data at scale while significantly reducing computational demands. JQL distills LLMs' annotation capabilities into lightweight annotators based on pretrained multilingual embeddings. These models exhibit robust multilingual and cross-lingual performance, even for languages and scripts unseen during training. Evaluated empirically across 35 languages, the resulting annotation pipeline substantially outperforms current heuristic filtering methods like Fineweb2. JQL notably enhances downstream model training quality and increases data retention rates. Our research provides practical insights and valuable resources for multilingual data curation, raising the standards of multilingual dataset development.