AI Paper Atlas

Research map / Label-efficient and robust learning / Knowledge distillation

Understanding the Effects of Projectors in Knowledge Distillation

Yudong Chen, Sen Wang, Jiajun Liu, Xuwei Xu, Frank de Hoog, Branislav Kusý, Zi Helen Huang

TPAMI 2026 · Label-efficient and robust learning › Knowledge distillation · approach: Distillation with causal and adaptive guidance

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add a single projector to feature distillation (ablation)feature distillation ResNet32x4-ResNet8x4 · CIFAR-100top-1 accuracy: 73.66 -> 75.14Improved
ensemble three projectors instead of one (ablation)projector ensemble ResNet32x4-ResNet8x4 · CIFAR-100top-1 accuracy: 75.14 -> 76.08Improved
ensemble four projectors instead of three (ablation)projector ensemble ResNet32x4-ResNet8x4 · CIFAR-100top-1 accuracy: 76.08 -> 75.93No clear effect
deepen projector to 4-layer MLP (ablation)projector ResNet32x4-ResNet8x4 · CIFAR-100top-1 accuracy: 75.14 -> 74.30Hurt
deepen projector to 4-layer MLP (ablation)projector VGG13-VGG8 · CIFAR-100top-1 accuracy: 73.84 -> 72.73Hurt
widen hidden layers of 2-layer MLP projector (ablation)2L-MLP projector VGG13-VGG8 · CIFAR-100top-1 accuracy: 73.31 -> 73.37No clear effect
add ReLU activation to projector ensemble (ablation)projector ensemble ResNet32x4-ResNet8x4 · CIFAR-100top-1 accuracy: 75.04 -> 76.08Improved
replace ReLU with GELU (ablation)projector ensemble ResNet32x4-ResNet8x4 · CIFAR-100top-1 accuracy: 76.08 -> 76.32Improved
use mixed instead of default initialization (ablation)projector ensemble ResNet32x4-ResNet8x4 · CIFAR-100top-1 accuracy: 76.08 -> 76.27Improved
use orthogonal instead of default initialization (ablation)projector ensemble VGG13-VGG8 · CIFAR-100top-1 accuracy: 74.35 -> 73.89Hurt
add projector to feature distillation on ImageNet (ablation)projector ResNet34-ResNet18 · ImageNettop-1 accuracy: 71.63 -> 71.94Improved
ensemble projectors for ResNet50-MobileNet (ablation)projector ensemble · ImageNettop-1 accuracy: 72.75 -> 73.29Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy