ewgf
@diwfaw
Vision-language models struggle with words like 'no' or 'not' in medical image analysis. This means they might mess up when asked to find images with specific objects but exclude others. Their lack of negation understanding can lead to unexpected failures.
0 reply
0 recast
0 reaction