It takes a pair of sentences as input and classifies the input pair to 'entailment' or 'no-entailment'. The class label entailment implies the second sentence entails the first sentence, and the no-entailment implies it does not. The Text Embedding model which is pre-trained on English and German Wikipedia returns an embedding of the input pair of sentences. The model available for deployment is created by attaching a binary classification layer to the output of the Text Embedding model, and then fine-tuning the entire model on [QNLI]