AI Paper Atlas

Research map / Federated learning and privacy / Differential privacy

Privacy-Preserving Model Transcription With Differentially Private Synthetic Distillation

Bochao Liu, Shiming Ge, Pengju Wang, Shikun Li, Tongliang Liu

TPAMI 2026 · Federated learning and privacy › Differential privacy · approach: Differentially private synthetic data · 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
add information entropy loss (ablation)DPSD generator · MNIST, data-sensitive, eps=10test accuracy: 66.37 -> 97.85Improved
add information entropy loss (ablation)DPSD generator · FMNIST, data-sensitive, eps=10test accuracy: 62.37 -> 90.08Improved
add cross entropy loss (ablation)DPSD generator · MNIST, data-sensitive, eps=10test accuracy: 96.02 -> 97.85Improved
add cross entropy loss (ablation)DPSD generator · FMNIST, data-sensitive, eps=10test accuracy: 79.46 -> 90.08Improved
add normalization loss (ablation)DPSD generator · MNIST, data-sensitive, eps=10test accuracy: 86.39 -> 97.85Improved
add normalization loss (ablation)DPSD generator · FMNIST, data-sensitive, eps=10test accuracy: 83.00 -> 90.08Improved
use student instead of teacher as discriminator (ablation)DPSD · MNISTtest accuracy: 97.93 -> 95.85Mixed
use student instead of teacher as discriminator (ablation)DPSD · FMNISTtest accuracy: 91.24 -> 90.08Mixed
use decoupled knowledge distillation instead of classic KD (ablation)DPSD · MNIST and FMNISTtest accuracyImproved
use normalization instead of clipping of gradients (ablation)DPSD annotation · label-sensitive, small boundtest accuracyImproved
increase iteration number T from 40 to 320 (ablation)DPSD · CIFAR100 and otherstest accuracyImproved
increase noise scale sigma from 20 to 100 then beyond (ablation)DPSD annotation · label-sensitivetest accuracyMixed

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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