AI Paper Atlas

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

SGD-Based Knowledge Distillation with Bayesian Teachers: Theory and Guidelines

Itai Morad, Nir Shlezinger, Yonina C. Eldar

ICLR 2026 · Label-efficient and robust learning › Knowledge distillation · approach: Distillation with causal and adaptive guidance · 0 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
use a Bayesian (VI) teacher instead of a deterministic teacherKD student · CIFAR-100 ResNet-50 -> WRN-16-2student accuracy (%): 69.36 -> 73.63 (+4.27)Improved
use a Bayesian teacher instead of deterministicKD student · CIFAR-100 VGG-13 -> WRN-40-1student accuracy (%): 67.90 -> 71.37 (+3.47)Improved
use a Bayesian teacher instead of deterministicKD student · CIFAR-100 WRN-40-2 -> WRN-16-2student accuracy (%): 70.80 -> 72.94 (+2.14)Improved
use a Bayesian teacher instead of deterministicKD student · CIFAR-100 ResNet-18 -> ResNet-18student accuracy (%): 75.92 -> 76.92 (+1.00)Improved
convert deterministic teacher to Bayesian via Laplace approximationKD student · CIFAR-100 ResNet-50 -> WRN-16-2student accuracy (%): 69.36 -> 72.52 (+3.16)Improved
convert deterministic teacher via Laplace approximationKD student · CIFAR-100 ResNet-50 -> ResNet-34student accuracy (%): 76.82 -> 76.23 (-0.59)Hurt
Laplace approximation teacherteacher accuracy · CIFAR-100 ResNet-50 teacherteacher accuracy (%): 75.41 -> 74.83 (-0.58)Hurt
train teacher with MSE loss instead of CEKD student · CIFAR-100 ResNet-50 -> WRN-16-2student accuracy (%): 69.36 -> 68.55 (-0.81)Hurt
fine-tune teacher with MCMIKD student · CIFAR-100 ResNet-50 -> WRN-16-2student accuracy (%): 69.36 -> 69.58 (+0.22)No clear effect
test-time data augmentation teacher (TTDA)KD student · CIFAR-100 ResNet-50 -> WRN-16-2student accuracy (%): 69.36 -> 69.24 (-0.12)No clear effect
distill from Bayesian teacher for stabilityKD student · CIFAR-100 ResNet-50 -> ResNet-18convergence noise: 30% less noiseImproved
supervise with true class probabilities instead of one-hot labelsSGD student · synthetic experimentloss neighborhoodImproved

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