Projected gradient matching reveals that medical image embeddings in vision transformers can be shifted between disease classes with imperceptible pixel changes, undermining the claim that these representations are semantically grounded.
Attention is all you need,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CV 1years
2025 1verdicts
REJECT 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
Are Vision Transformer Representations Semantically Meaningful? A Case Study in Medical Imaging
Projected gradient matching reveals that medical image embeddings in vision transformers can be shifted between disease classes with imperceptible pixel changes, undermining the claim that these representations are semantically grounded.