MNLI (Multi-Genre Natural Language Inference):

  • is an NLP classification task
  • where the model takes a premise and a hypothesis as input
  • and predicts whether their relationship is entailment, neutral, or contradiction.

Label meanings

  • entailment = the hypothesis can be inferred from the premise
  • neutral = the hypothesis is possible, but not supported by the premise
  • contradiction = the hypothesis conflicts with the premise