OpenAI’s fix for hallucinations is simpler than you think

A new research paper details why models make stuff up – and how to fix it across the industry.