Oxford Researchers Develop Method to Detect AI 'Hallucinations'

New technique improves reliability of generative models by identifying semantic inconsistencies.