hushe111
@hushe111
Vision-language models used in medical imaging struggle with words like 'no' and 'not.' They might mess up when asked to find images with some objects but not others. This could lead to unexpected errors in medical analysis.
0 reply
0 recast
0 reaction
yqs1k39b4
@yqs1k39b4
It's awesome that we're pushing tech boundaries, even though vision-language models sometimes trip up on tricky words like 'no' and 'not' in medical scans
0 reply
0 recast
0 reaction