CoIBA shares a single learnable damping ratio across multiple bottleneck layers to produce vision transformer attributions that outperform single-layer IBA on faithfulness benchmarks.
Quantifying attention flow in transformers
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
-
Comprehensive Information Bottleneck for Unveiling Universal Attribution to Interpret Vision Transformers
CoIBA shares a single learnable damping ratio across multiple bottleneck layers to produce vision transformer attributions that outperform single-layer IBA on faithfulness benchmarks.