AI Paper Atlas

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

Self-Distillation as Instance-Specific Label Smoothing

Zhilu Zhang, Mert Sabuncu

NeurIPS 2020 · Label-efficient and robust learning › Knowledge distillation · approach: Rethinking distillation objectives · 63 citations (OpenAlex)

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
Increase teacher temperatureSelf-distillation with ResNet-34 · CIFAR-100test accuracyImproved
Increase teacher temperature beyond the point where confidence diversity plateausSelf-distillation with ResNet-34 · CIFAR-100NLLHurt
Use self-distillation instead of label smoothing at matched average smoothingResNet and DenseNet · CIFAR-100, CUB-200, and Tiny-ImageNettest accuracyImproved
Use Beta smoothing instead of label smoothingResNet and DenseNet · CUB-200 and Tiny-ImageNet; CIFAR-100 ResNet is an exceptiontest accuracyMixed
Use self-distillation instead of label smoothingResNet and DenseNet · CIFAR-100, CUB-200, and Tiny-ImageNetexpected calibration errorImproved
Use Beta smoothing instead of label smoothingResNet and DenseNet · CIFAR-100, CUB-200, and Tiny-ImageNetexpected calibration errorImproved
Use EMA confidence rankings to assign Beta-smoothed labels rather than assigning random Beta labels (ablation)Beta smoothing · CIFAR-100, CUB-200, and Tiny-ImageNettest accuracyImproved
Use EMA confidence rankings to assign Beta-smoothed labels rather than assigning random Beta labels (ablation)Beta smoothing · CIFAR-100, CUB-200, and Tiny-ImageNetexpected calibration errorImproved
Use Beta smoothing instead of explicit self-training with EMA predictions (ablation)ResNet · CIFAR-100, CUB-200, and Tiny-ImageNettest accuracyImproved
Use Beta smoothing instead of explicit self-training with EMA predictions (ablation)ResNet · CIFAR-100, CUB-200, and Tiny-ImageNetexpected calibration errorImproved
Use a better teacher to provide instance-specific priors (ablation)Cross-distillation · CIFAR-100, CUB-200, and Tiny-ImageNettest accuracyImproved
Prune 50% of classes from teacher predictions during distillation (ablation)Pruned distillation · CIFAR-100 and Tiny-ImageNettest accuracyNo clear effect

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