Glossary term
Glossary term
Foundations
NLP task of determining whether a hypothesis is entailed by, contradicted by, or neutral with respect to a premise.
Facebook's Bart-large-MNLI model is used as a zero-shot text classifier by thousands of developers on Hugging Face, enabling topic classification without task-specific training by treating labels as hypotheses.
Google Search uses NLI models to assess factual consistency between retrieved passages and generated answers in AI Overviews, flagging cases where the summary contradicts the source material.
Snopes.com's fact-checking pipeline uses NLI to automatically flag candidate claims that contradict a known corpus of verified facts, surfacing high-priority items for human fact-checkers to review.