# Cross-lingual Adaptation

Arabic CLIP is an adapted version of the Contrastive Language-Image Pre-training (CLIP) model for Arabic, capable of learning concepts from images and associating them with Arabic text descriptions.

This model is a dense retrieval model further fine-tuned on all Mr. TyDi training data based on the castorini/mdpr-tied-pft-msmarco checkpoint.

AfroXLMR-mini is created by adapting the XLM-R-miniLM model through masked language model (MLM) training on 17 African languages, covering major African language families and three high-resource languages (Arabic, French, and English).