AI Paper Atlas

Research map / Optimization / Implicit bias and generalization

Stabilizing Sharpness-Aware Minimization Through A Simple Renormalization Strategy

Chengli Tan, Jiangshe Zhang, Junmin Liu, Yicheng Wang, Yunda Hao

JMLR 2025 · Optimization › Implicit bias and generalization · approach: Sharpness-aware minimization theory · 3 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 gradient renormalization (SSAM)SAM · ImageNet-1K val, ViT-S-32top-1 accuracy (%): 69.98 -> 71.15Improved
add gradient renormalization (SSAM)SAM · ImageNet-1K val, ViT-S-16top-1 accuracy (%): 76.88 -> 77.41Improved
add gradient renormalization (SSAM)SAM · ImageNet-1K val, ResNet-18top-1 accuracy (%): 70.66 -> 70.76Improved
add gradient renormalization (SSAM)SAM · ImageNet-1K val, ResNet-50top-1 accuracy (%): 77.82 -> 77.89Improved
add gradient renormalization (SSAM)SAM · CIFAR-10, PyramidNet-110test accuracy (%): 96.91 -> 97.04Improved
add gradient renormalization (SSAM)SAM · CIFAR-100 (5 backbones)test accuracy (%): ResNet-20 70.77 -> 70.48; WRN-28-10 82.94 -> 82.80Mixed
add gradient renormalization (SSAM)SAM · Quartic saddle toy function, large learning raterate of successful trainingImproved
add gradient renormalization (SSAM)SAM · Quadratic loss d=20convergence speedHurt
add gradient renormalization (SSAM)SAM · ResNet-20 CIFAR-10 / ResNet-56 CIFAR-100top Hessian eigenvalue (sharpness)Improved
add gradient norm normalization in SAM (SAM* vs SAM)SAM · CIFAR-10 and CIFAR-100test accuracy (%): CIFAR-10 WRN-28-10 96.55 -> 96.91No clear effect
add sharpness-aware ascent step (SAM instead of SGD)SGD · CIFAR-10, ResNet-20test accuracy (%): 92.78 -> 93.43Improved
use GASAM instead of SAMSAM · CIFAR-10, PyramidNet-110test accuracy (%): 96.91 -> 81.83Hurt

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