AI Paper Atlas

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

Good Teachers Explain: Explanation-Enhanced Knowledge Distillation

Amin Parchami-Araghi, Moritz Böhle, Sukrut Rao, Bernt Schiele

ECCV 2024 · Label-efficient and robust learning › Knowledge distillation · approach: Distillation with causal and adaptive guidance · 21 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
add explanation-matching loss (e2KD, GradCAM) to logit KD, 50 shotsKD, ResNet-34 -> ResNet-18 · ImageNet val, 50 shotstop-1 accuracy (%): 49.8 -> 54.9Improved
add e2KD to KD with full dataKD, ResNet-34 -> ResNet-18 · ImageNet val, full datatop-1 accuracy (%): 71.8 -> 71.8No clear effect
add e2KD (B-cos) to KDKD, B-cos ResNet-34 -> B-cos ResNet-18 · ImageNet val, 50 shotsteacher agreement (%): 38.4 -> 48.4Improved
add e2KD to CRDCRD · ImageNet val, 200 shotstop-1 accuracy (%): 51.0 -> 54.1Improved
add e2KD across different student architecturesKD on Waterbirds-100 · Waterbirds-100 out-of-distributionaccuracy (%): ConvNeXt 20.5 -> 32.2; ShuffleNet 23.1 -> 37.0Improved
add e2KD so the student matches a guided teacher's explanationsKD on PASCAL VOC · PASCAL VOC EPG teacherF1: 60.1 -> 67.6Improved
distill CNN teacher to ViT student with e2KDB-cos ViT-Tiny student · ImageNet valteacher agreement (%): 70.1 -> 71.8Improved
use frozen pre-computed explanations with e2KD instead of frozen-logit KD (ablation)B-cos DenseNet-169 -> ResNet-18, 50 shots · ImageNet valtop-1 accuracy (%): 33.4 -> 38.7Improved
remove 2x2 down-sampling of CAMs in CAT-KD without rescaling the loss (ablation)CAT-KD · ImageNet 50 shotstop-1 accuracy (%): 32.2 -> 24.4Hurt
replace CE with KL-divergence and use top-1 explanation in CAT-KD (ablation)CAT-KD · ImageNet 50 shotstop-1 accuracy (%): 35.0 -> 53.1Improved
replace CAM with GradCAM in the CAT-KD variant (ablation)CAT-KD variant · ImageNet 50 shotstop-1 accuracy (%): 53.1 -> 52.9No clear effect
longer teaching schedule and mixup for KD (ablation)KD on Waterbirds-100 · Waterbirds-100agreementImproved

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