AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Large Learning Rates Simultaneously Achieve Robustness to Spurious Correlations and Compressibility

Melih Barsbey, Lucas Prieto, Stefanos Zafeiriou, Tolga Birdal

ICCV 2025 · Label-efficient and robust learning › Recognition and long-tail · approach: Group-robust classification · 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
increase learning rate across the sweep range (ablation)ResNet18 · Colored MNISTunbiased test accuracy (%): 68.343 -> 91.543Improved
increase learning rate across the sweep range (ablation)ResNet50 · Double MNISTunbiased test accuracy (%): 42.045 -> 95.733Improved
increase learning rate across the sweep range (ablation)Wide ResNet101-2 · Double MNISTunbiased test accuracy (%): 43.12 -> 96.08Improved
increase learning rate across the sweep range (ablation)ResNet18 · MNIST-CIFARunbiased test accuracy (%): 26.233 -> 47.513Improved
increase learning rate across the sweep range (ablation)Swin Transformer · CelebA (worst-group)worst-group accuracy (%): 40.889 -> 48.21Improved
increase learning rate across the sweep range (ablation)FCN · Colored MNISTunbiased test accuracy (%): 72.727 -> 76.11Improved
increase learning rate with PSGD optimizer (ablation)ResNet50 (PSGD) · Colored MNISTunbiased test accuracy (%): 24.16 -> 93.09Improved
add L2 regularization on top of high learning rate (ablation)ResNet18 · Colored MNISTmax unbiased test accuracy (%): 92.203 -> 94.25Improved
add L2 regularization on top of high learning rate (ablation)ResNet18 · Double MNISTmax unbiased test accuracy (%): 96.39 -> 97.75Improved
add L2 regularization on top of high learning rate (ablation)ResNet18 · MNIST-CIFARmax unbiased test accuracy (%): 47.497 -> 48.327Improved
use ASAM rho sweep instead of learning rate sweep (ablation)ResNet18 · MNIST-CIFARmax unbiased test accuracy (%): 47.497 -> 34.053Hurt
use focal loss gamma sweep instead of learning rate sweep (ablation)ResNet18 · MNIST-CIFARmax unbiased test accuracy (%): 47.497 -> 40.393Hurt

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