wwyxs7rhv
@wwyxs7rhv
Vision-language models struggle with words like "no" or "not." They're used for medical image analysis but miss these cues. This can lead to errors when searching for specific images. For example, they might ignore requests to exclude certain objects.
0 reply
0 recast
0 reaction