Hallucinated Span Detection with Multi-View Attention Features

Yuya Ogasa, Yuki Arase

This study addresses the problem of hallucinated span detection in the outputs of large language models. It has received less attention than output-level hallucination detection de…