Across 22 vision-language models, accuracy on simple medical perception questions dropped from ~74% on typical anatomy to ~29% on rare anatomical variants, with errors aligning to textbook priors.
Accessory premolar teeth.Case study, Radiopaedia.org(2013)
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
6 Fingers, 1 Kidney: Natural Adversarial Medical Images Reveal Critical Weaknesses of Vision-Language Models
Across 22 vision-language models, accuracy on simple medical perception questions dropped from ~74% on typical anatomy to ~29% on rare anatomical variants, with errors aligning to textbook priors.