1 paper · 1 filter
Shota Nakada, Kazuhiro Saito, Yuchi Ishikawa +3
We propose a novel task, hallucination localization in video captioning, which aims to identify hallucinations in video captions at the span level (i.e. individual words or phrases…