AI RESEARCH

A better method for identifying overconfident large language models

MIT AI News

This new metric for measuring uncertainty could flag hallucinations and help users know whether to trust an AI model.