AI Paper Atlas

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

Generalized Kullback–Leibler Divergence Loss

Jiequan Cui, Beier Zhu, Qingshan Xu, Zhuotao Tian, Xiaojuan Qi, Bei Yu, Hanwang Zhang, Richang Hong

TPAMI 2026 · Label-efficient and robust learning › Knowledge distillation · approach: Few-shot and metric distillation · 25 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
break asymmetric optimization (BA) by enabling gradient on wMSE (ablation)DKL loss · CIFAR-100 AT / ImageNet KDAA / top-1: 30.20 -> 30.42 AA; 71.03 -> 71.60 top-1Improved
smoother sample-wise weight function (gamma 0.4) (ablation)DKL with BA · CIFAR-100 ATclean / AA: 62.69 -> 64.15 clean; 30.42 -> 31.70 AAImproved
class-wise weight function instead of sample-wise (ablation)DKL with BA · CIFAR-100 ATclean / AA: 64.15 -> 65.76 clean; 31.70 -> 31.91 AAImproved
class-wise smoother weight function (ablation)DKL with BA · ImageNet KDtop-1: 71.80 -> 71.91Improved
replace KL with DKL (equivalent form) (ablation)KL baseline · CIFAR-100 ATAA: 30.29 -> 30.20No clear effect
GKL instead of KL, temperature 1.0KL-KD · ImageNet-LTAll accuracy: 44.32 -> 45.40Improved
raise KD temperature from 1.0 to 2.0 (ablation)KL-KD · ImageNet-LTMany / Few accuracy: 64.60 -> 66.50 Many; 9.53 -> 6.54 FewMixed
GKL-KD instead of KL-KD (ResNeXt-101 to ResNet-50)KL-KD · ImageNet-LTMany / Medium accuracy: 68.83 -> 70.31 Many; 42.31 -> 43.47 MediumImproved
class-wise global information with gamma 0.3 (ablation)KD weight function · ImageNet-LTaccuracy: 44.32 -> 45.40Improved
larger KD temperature t=1 to t=8 (ablation)KL-KD CLIP · ImageNet zero-shottop-1: 57.28 -> 60.85Improved
GKL loss instead of KLFixMatch semi-supervised · CIFAR-100top-1: 69.09 -> 70.00Improved
GKL loss instead of MSE consistencyMean-Teacher · CIFAR-100top-1: 67.38 -> 70.05Improved

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