Omartificial-Intelligence-Space/Arab3M-Triplets

Arab3M-Triplets This dataset is designed for training and evaluating models using contrastive learning techniques, particularly in the context of natural language understanding. The dataset consists of triplets: an anchor sentence, a positive sentence, and a negative sentence. The goal is to encourage models to learn meaningful representations by distinguishing between semantically similar and dissimilar sentences. Dataset Overview Format: Parquet Number of rows:… See the full description on the dataset page: https://huggingface.co/datasets/Omartificial-Intelligence-Space/Arab3M-Triplets.