AI Paper Atlas

Research map / Architectures and efficiency / Pruning and efficient tuning

On the Robustness Tradeoff in Fine-Tuning

Kunyang Li, Jean-Charles Noirot Ferrand, Ryan Sheatsley, Blaine Hoak, Yohan Beugin, Eric Pauley, Patrick McDaniel

ICCV 2025 · Architectures and efficiency › Pruning and efficient tuning · approach: Adapter-based parameter-efficient tuning · 9 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
fine-tune longer after early stepsViT-B fine-tuning · Caltech256adversarial robustness: 25% -> 10%Hurt
fine-tune the first stepsViT-B fine-tuning · Caltech256accuracy and robustness: accuracy ≈ 90% within 1,000 stepsImproved
use BitFit instead of average methodViT-B fine-tuning · CIFAR-10Pareto AUC: 75% above averageImproved
use BitFit instead of average methodViT-B fine-tuning · CIFAR-100Pareto AUC: 81.5%Improved
use Compacter instead of average methodViT-B fine-tuning · Caltech256Pareto AUC: 57.5% above averageImproved
use Compacter instead of average methodViT-B fine-tuning · CUB-200Pareto AUC: 34.6% above averageImproved
use Compacter instead of average methodViT-B fine-tuning · Stanford DogsPareto AUC: 24% above averageImproved
use linear probing rather than other methodsViT-B fine-tuning · CIFAR-100Pareto AUC: 47.5% lower than averageHurt
use full fine-tuning rather than PEFTViT-B fine-tuning · CIFAR-100Pareto AUC: 20% lower than averageHurt
use full fine-tuning rather than linear probingViT-B fine-tuning · DomainNet OODOOD robustness: 61% -> 73%Improved
train on a more shifted domain instead of realViT-B fine-tuning · DomainNet OODOOD robustness: 64% -> 73%Improved
vary LoRA rank (1-20), adapter reduction factor (4-32) or update location (ablation)PEFT · Adversarial robustnessadversarial robustnessNo clear effect

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