AI Paper Atlas

Research map / Federated learning and privacy / Differential privacy

Private and Stable Test-time Adaptation with Differential Privacy

Zefeng Li, Qiaoyue Tang, Mathias Lécuyer, Evan Shelhamer

ICML 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 per-sample gradient clipping (no noise) to TentTent · ImageNet-C sev 5, continual, ViT-B/16avg top-1 accuracy (%): 60.8 -> 64.9Improved
add per-sample gradient clipping to EATAEATA · ImageNet-C sev 5, continual, ViT-B/16avg top-1 accuracy (%): 65.9 -> 66.0No clear effect
add per-sample gradient clipping to SARSAR · ImageNet-C sev 5, continual, ViT-B/16avg top-1 accuracy (%): 60.8 -> 61.7Improved
add per-sample gradient clipping to DeYODeYO · ImageNet-C sev 5, continual, ViT-B/16avg top-1 accuracy (%): 61.0 -> 62.7Improved
add per-sample gradient clipping to DeYO-COMEDeYO-COME · ImageNet-C sev 5, continual, ViT-B/16avg top-1 accuracy (%): 66.4 -> 67.5Improved
make Tent differentially private at epsilon=10 to 20Tent · ImageNet-C sev 5, continual, ViT-B/16avg top-1 accuracy (%): 60.8 -> 62.9 (epsilon=20)Improved
make EATA, SAR, DeYO, DeYO-COME private at epsilon=20EATA, SAR, DeYO, DeYO-COME · ImageNet-C sev 5, continual, ViT-B/16avg top-1 accuracy (%): -2.9%, -1.2%, -2.4%, -1.7%Hurt
strengthen privacy to epsilon=1 for DP-Tent (ablation)DP-Tent · ImageNet-C sev 5, continual, ViT-B/16avg top-1 accuracy (%): accuracy degrades by 2.3%Hurt
use batch-level clipping instead of per-sample clipping (ablation)Tent · ImageNet-C sev 5, continual, ViT-B/16avg top-1 accuracy (%): 60.8 -> 59.8 (C=1; Tent in Table 2 59.7)No clear effect
replace hard filtering by grad norm with rescaling by per-sample clipping (ablation)Tent · ImageNet-C sev 5, continual, ViT-B/16avg top-1 accuracy (%): 55.0 -> 64.9 at C=1Improved
filter by entropy loss instead of clipping (ablation)Tent · ImageNet-C sev 5, continual, ViT-B/16avg top-1 accuracy (%): 64.9 -> 62.9Hurt
add EATA filters (Filter-1 and Filter-2) to Tent (ablation)Tent · ImageNet-C sev 5, episodic and continualavg top-1 accuracy (%): episodic 63.0 -> 65.5, continual 60.8 -> 63.7Improved

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