DIDB-ViT combines differential attention, Haar-wavelet frequency decomposition, and token-wise activation shifts to improve binary vision transformers, achieving state-of-the-art results on several benchmarks.
Improving vision transformers by revisit- ing high-frequency components
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
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
High-Fidelity Differential-information Driven Binary Vision Transformer
DIDB-ViT combines differential attention, Haar-wavelet frequency decomposition, and token-wise activation shifts to improve binary vision transformers, achieving state-of-the-art results on several benchmarks.