AI Paper Atlas

Research map / Label-efficient and robust learning / Self-supervised and contrastive

"Trust, but Verify: Using Self-Supervised Probing to Improve Trustworthiness"

Ailin Deng, Shen Li, Miao Xiong, Zhirui Chen, Bryan Hooi

ECCV 2022 · Label-efficient and robust learning › Self-supervised and contrastive · approach: Data-efficient and robust self-supervision · 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 self-supervised probing confidence to MSPMSP · CIFAR-10 misclassification (ResNet-18)AUROC: 90.63 -> 92.10Improved
add self-supervised probing to MCDropoutMCDropout · CIFAR-10 misclassification (ResNet-18)AUPR-ERR: 42.04 -> 52.60Improved
add self-supervised probing to TrustScoreTrustScore · CIFAR-10 misclassification (VGG16)AUROC: 90.47 -> 91.68Improved
add self-supervised probing to MSPMSP · CINIC-10 misclassification (VGG16)AUROC: 86.33 -> 85.53Hurt
add self-supervised probing to TCPTCP · CIFAR-10 misclassification (ResNet-18)AUROC: 93.74 -> 93.73No clear effect
add translation probing task to rotation probing (ablation)SSP · CIFAR-10 misclassification (ResNet-18)AUROC: 91.78 -> 92.10Improved
add rotation probing task to translation probing (ablation)SSP · CIFAR-10 misclassification (ResNet-18)AUROC: 91.42 -> 92.10Improved
add SSP to MSP for OOD detectionMSP · CIFAR-10 to SVHN OOD (VGG16)AUROC: 91.05 -> 92.57Improved
add SSP to Entropy for OOD detectionEntropy · CIFAR-10 to LSUN OOD (ResNet-18)AUROC: 91.94 -> 93.55Improved
add SSP to MSP for OOD detectionMSP · CINIC-10 to SVHN OOD (VGG16)AUROC: 81.48 -> 85.25Improved
add input-dependent temperature from probing to temperature scalingTemp. Scaling · CIFAR-10 calibration (ResNet-18)ECE: 1.40 -> 0.75Improved
add input-dependent temperature from probing to temperature scalingTemp. Scaling · CIFAR-10 calibration (VGG16)MCE: 7.53 -> 9.15Hurt

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