AI Paper Atlas

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

Better May Not Be Fairer: A Study on Subgroup Discrepancy in Image Classification

Ming-Chang Chiu, Pin-Yu Chen, Xuezhe Ma

ICCV 2023 · Label-efficient and robust learning › Recognition and long-tail · approach: Correspondence and causal representation learning · 5 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
apply FlowAug flow-latent augmentation instead of Standard trainingResNet18 · CIFAR-10accuracy, best: 95.16 -> 96.58Improved
apply FlowAug instead of Standard trainingResNet18 · CIFAR-100accuracy, best: 78.52 -> 79.99Improved
apply FlowAug instead of Standard trainingResNet18 · CIFAR10-BMacroStd: 2.24 -> 1.65Improved
perturb or mix the local representation nu instead of global z (ablation)FlowAug · CIFAR-10 / CIFAR-100accuracy, best: CIFAR-10 96.58 -> 94.22; CIFAR-100 79.99 -> 74.05Hurt
use uniform noise instead of truncated Gaussian for z perturbation (ablation)FlowAug · CIFAR-10accuracy, best: 96.23 -> 96.12No clear effect
combine FlowAug with Cutmix (ablation)FlowAug · CIFAR-100accuracy: 79.45 -> 81.57Improved
add Standard loss and Mix z to FlowAug with uniform noise (ablation)FlowAug · CIFAR-10accuracy, best: 96.12 -> 96.58Improved
apply Mixup and Cutmix on CIFAR-100 subgroup discrepancyResNet18 · CIFAR100-BMacroStd: 12.24 -> 12.62 (Cutmix)Hurt
use MacroStd instead of WeightedStd as sensitivity metricmetric · CIFAR-10correlation with accuracy: -0.78 -> -0.89Improved
apply FlowAug on ImageNet-10ResNet18 · ImageNet-10accuracy: 91.80 -> 94.80Improved

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