A simple redundancy-suppression loss, applied to penultimate-layer embeddings, outperforms the complex OFA method for cross-architecture knowledge distillation while adding far fewer parameters.
Knowledge distillation with the reused teacher classifier
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
background 1representative citing papers
citing papers explorer
-
Cross-Architecture Distillation Made Simple with Redundancy Suppression
A simple redundancy-suppression loss, applied to penultimate-layer embeddings, outperforms the complex OFA method for cross-architecture knowledge distillation while adding far fewer parameters.