Cross-Architecture Distillation Made Simple with Redundancy Suppression

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhang, Weijia, Liu, Yuehao, Ran, Wu, Ma, Chao
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911081761341440
author Zhang, Weijia
Liu, Yuehao
Ran, Wu
Ma, Chao
author_facet Zhang, Weijia
Liu, Yuehao
Ran, Wu
Ma, Chao
contents We describe a simple method for cross-architecture knowledge distillation, where the knowledge transfer is cast into a redundant information suppression formulation. Existing methods introduce sophisticated modules, architecture-tailored designs, and excessive parameters, which impair their efficiency and applicability. We propose to extract the architecture-agnostic knowledge in heterogeneous representations by reducing the redundant architecture-exclusive information. To this end, we present a simple redundancy suppression distillation (RSD) loss, which comprises cross-architecture invariance maximisation and feature decorrelation objectives. To prevent the student from entirely losing its architecture-specific capabilities, we further design a lightweight module that decouples the RSD objective from the student's internal representations. Our method is devoid of the architecture-specific designs and complex operations in the pioneering method of OFA. It outperforms OFA on CIFAR-100 and ImageNet-1k benchmarks with only a fraction of their parameter overhead, which highlights its potential as a simple and strong baseline to the cross-architecture distillation community.
format Preprint
id arxiv_https___arxiv_org_abs_2507_21844
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Cross-Architecture Distillation Made Simple with Redundancy Suppression
Zhang, Weijia
Liu, Yuehao
Ran, Wu
Ma, Chao
Computer Vision and Pattern Recognition
We describe a simple method for cross-architecture knowledge distillation, where the knowledge transfer is cast into a redundant information suppression formulation. Existing methods introduce sophisticated modules, architecture-tailored designs, and excessive parameters, which impair their efficiency and applicability. We propose to extract the architecture-agnostic knowledge in heterogeneous representations by reducing the redundant architecture-exclusive information. To this end, we present a simple redundancy suppression distillation (RSD) loss, which comprises cross-architecture invariance maximisation and feature decorrelation objectives. To prevent the student from entirely losing its architecture-specific capabilities, we further design a lightweight module that decouples the RSD objective from the student's internal representations. Our method is devoid of the architecture-specific designs and complex operations in the pioneering method of OFA. It outperforms OFA on CIFAR-100 and ImageNet-1k benchmarks with only a fraction of their parameter overhead, which highlights its potential as a simple and strong baseline to the cross-architecture distillation community.
title Cross-Architecture Distillation Made Simple with Redundancy Suppression
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2507.21844