Post by Los Alamos National Laboratory

Seeing or hallucinating? 👁️ Los Alamos researchers developed a new tool that helps vision-language AI models identify when their responses may not be grounded in the image they're analyzing. The approach, called Prelim Attention Score (PAS), could help improve the reliability of AI systems used for scientific analysis, engineering applications and other mission-relevant workflows where accuracy matters. See the science 👉 https://ow.ly/yNVS50Zbl95