Leveraging LaBSE with Progressive Curriculum Learning for Multicultural Polarization

The article introduces a novel architecture for detecting online polarization in multilingual and multicultural contexts, utilizing LaBSE embeddings to improve cross-lingual learning, resulting in a macro F1 score increase of up to 0.2 in low-resource languages. It also presents an ablation study on various encoder models from the Qwen model family within a retrieval-based prompting framework. This work is significant for practitioners as it addresses the challenge of data scarcity in low-resource languages, enhancing the capabilities of AI systems in understanding and mitigating online polarization.