DayNews.aiMultimodal AI ignores non-text clues
A new benchmark reveals multimodal AI models fixate on text, ignoring audio or visual evidence even when it contradicts the answer.
DayNews.aiDayNews.aiDayNews.aiA new benchmark reveals multimodal AI models fixate on text, ignoring audio or visual evidence even when it contradicts the answer.