ershiyige
@ershiyige
Vision-language models used in medical imaging struggle with words like 'no' and 'not.' They might mess up when asked to find images with specific objects but exclude others. This could lead to unexpected errors in analyzing medical scans. The models just don't get negation.
0 reply
0 recast
0 reaction