AI Paper Atlas

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

No Subclass Left Behind: Fine-Grained Robustness in Coarse-Grained Classification Problems

Nimit Sohoni, Jared Dunnmon, Geoffrey Angus, Albert Gu, Christopher Ré

NeurIPS 2020 · Label-efficient and robust learning › Recognition and long-tail · approach: Group-robust classification · 42 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
train with GEORGE (cluster-based GDRO) instead of ERMResNet · Waterbirdsrobust accuracy: 63.3 -> 76.2Improved
train with GEORGE instead of ERMResNet · CelebArobust accuracy: 41.1 -> 52.4Improved
train with GEORGE instead of ERMResNet · U-MNISTrobust accuracy: 93.9 -> 95.7Improved
train with GEORGE instead of ERMResNet · ISIC histopathology subclassrobust AUROC: .872 -> .881No clear effect
train with GEORGE instead of ERMResNet · ISIC non-patch subclassrobust AUROC: .920 -> .918No clear effect
train with GEORGE instead of ERMResNet · Waterbirdsoverall accuracy: 97.2 -> 95.5Hurt
train with GEORGE instead of ERMResNet · CelebAoverall accuracy: 95.7 -> 94.8Hurt
cluster BiT pretrained embeddings rather than ERM features in step 1 (ablation)GEORGE · CelebArobust accuracy: 52.4 -> 86.0Improved
cluster BiT pretrained embeddings rather than ERM features in step 1 (ablation)GEORGE · Waterbirdsrobust accuracyHurt
use true subclass labels for GDRO instead of estimated clusters (ablation)GEORGE · CelebArobust accuracy: 52.4 -> 85.9Improved
use true subclass labels for GDRO instead of estimated clusters (ablation)GEORGE · Waterbirdsrobust accuracy: 76.2 -> 90.7Improved
apply UMAP dimensionality reduction before clustering (ablation)GEORGE · hidden stratification datasetsrobust accuracyImproved

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