AI Paper Atlas

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

A framework for benchmarking Class-out-of-distribution detection and its application to ImageNet

Ido Galil, Mohammed Dabbah, Ran El-Yaniv

ICLR 2023 · Label-efficient and robust learning › OOD and anomaly detection · approach: Spurious-aware OOD detection · 2 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 knowledge distillation during trainingImageNet classifiers with matched architectures and training regimes · ImageNet C-OOD, most severity levels above s3AUROCImproved
Pretrain on ImageNet-21kImageNet classifiers · ImageNet C-OOD, severity s≤3AUROCImproved
Add adversarial trainingImageNet classifiers · ImageNet C-OODAUROCNo clear effect
Replace softmax confidence with ODINImageNet classifiers · ImageNet C-OOD across severity levelsAUROC: 0.4%; 9%Mixed
Replace softmax confidence with max-logitImageNet classifiers · ImageNet C-OODAUROC: +10.6%Mixed
Replace softmax confidence with negative entropyImageNet classifiers · ImageNet C-OOD, most severity levelsAUROCImproved
Replace softmax confidence with negative entropyImageNet classifiers · ImageNet C-OOD, severity levels s∈{7,8,9}AUROCHurt
Add 30 dropout-enabled forward passes and predictive entropyImageNet classifiers · ImageNet C-OOD, especially low severityAUROCImproved
Add dropout-enabled sampling to entropy scoring (ablation)ImageNet classifiers using MC dropout · ImageNet C-OOD, most severity levelsAUROCImproved
Use CLIP cosine-similarity scores without training or fine-tuningCLIP ResNet-50x64 · ImageNet C-OOD, severity levels below 6AUROCImproved
Use CLIP cosine-similarity scoresCLIP ResNet-50x64 · ImageNet C-OOD, severity 10AUROCHurt

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