AI Paper Atlas

Research map / Label-efficient and robust learning / OOD and anomaly detection

Exploiting Local Flatness for Efficient Out-of-Distribution Detection

Seonghwan Park, Hyunji Jung, Dongyeop Lee, Namhoon Lee

ECCV 2026 · Label-efficient and robust learning › OOD and anomaly detection · approach: Spurious-aware OOD detection · 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 partial feature normalization (0<alpha<1) to the baselines and Fold (ablation)MSP, EBO, ReAct, ASH, Fold variants · CIFAR-10AUROC: 87.13 -> 91.13Improved
add partial feature normalization (0<alpha<1) to the baselines and Fold (ablation)MSP, EBO, ReAct, ASH, Fold variants · ImageNet-200AUROC: 87.64 -> 89.91Improved
use full normalization (alpha=1) instead of partial on ImageNet-200 (ablation)MSP, EBO, ReAct, ASH, Fold variants · ImageNet-200AUROC: 89.91 -> 79.02Hurt
use full normalization (alpha=1) instead of none on CIFAR-10 (ablation)Fold and baselines · CIFAR-10AUROC: 87.13 -> 91.34Improved
compute curvature in feature space instead of logit space (ablation)Fold · Four benchmarks averageAUROC: 86.02 -> 86.81Improved
use feature Hessian instead of logit Hessian on CIFAR-10 (ablation)Fold · CIFAR-10AUROC: 92.52 -> 92.46No clear effect
use feature Hessian instead of logit Hessian on ImageNet-200 (ablation)Fold · ImageNet-200AUROC: 86.35 -> 88.65Improved
add partial normalization, raising Cohen's d on ImageNet-200 (ablation)Fold · ImageNet-200Cohen's d: 1.605 -> 1.689Improved
full normalization collapses score separability (ablation)Fold · ImageNet-200Cohen's d: 1.605 -> 0.520Hurt
combine Fold with ReAct (Fold-r)Fold · Four benchmarks averageAUROC: 87.20 -> 87.91Improved
combine Fold with ASH (Fold-a) on CIFAR-10Fold · CIFAR-10AUROC: 92.46 -> 89.43Hurt
replace grid-searched alpha (OOD validation) with AutoFold ID logit-masking tuningFold · Four benchmarks averageAUROC: 87.20 -> 86.24Mixed

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