AI Paper Atlas

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

Rethinking Confidence Calibration for Failure Prediction

Fei Zhu, Zhen Cheng, Xu-Yao Zhang, Cheng-Lin Liu

ECCV 2022 · Label-efficient and robust learning › Recognition and long-tail · approach: Confidence calibration objectives · 46 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 mixupResNet110 · CIFAR-10AUROC: 90.76 -> 86.93Hurt
train with label smoothingResNet110 · CIFAR-10AUROC: 90.76 -> 81.89Hurt
train with label smoothingResNet110 · CIFAR-10AURC: 9.94 -> 24.50Hurt
train with mixupResNet110 · CIFAR-10ECE: 4.14 -> 2.97Improved
train with focal lossResNet110 · CIFAR-10ECE: 4.14 -> 1.60Improved
train with focal lossResNet110 · CIFAR-10AUROC: 90.76 -> 90.70No clear effect
add temperature scalingclassifiers · CIFAR/ImageNetfailure predictionNo clear effect
add SAM (flat minima) to trainingResNet110 · CIFAR-10AUROC: 90.76 -> 93.73Improved
add SWA (flat minima) to trainingResNet110 · CIFAR-10AUROC: 90.76 -> 93.69Improved
combine SAM and SWA (FMFP)ResNet110 · CIFAR-10AUROC: 90.76 -> 93.76Improved
combine SAM and SWA (FMFP)ResNet110 · CIFAR-100AUROC: 85.00 -> 87.24Improved
use FMFP instead of CRLResNet110 · CIFAR-10AUROC: 92.86 -> 93.76Improved

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