The Arabic Version of SNLI and MultiNLI datasets, originally used for Natural Language Inference (NLI), may be used for finetuning embedding models.