SOTAVerified

Error Understanding

Discover what causes the model’s prediction errors.

Papers

Recently AddedMost HypedMost ActiveNeeds VerificationMost Verified

Showing 1–9 of 9 papers

TitleStatusHype
Less is More: Fewer Interpretable Region via Submodular Subset SelectionCode2
Making Sense of Dependence: Efficient Black-box Explanations Using Dependence MeasureCode1
Grad-CAM++: Improved Visual Explanations for Deep Convolutional NetworksCode1
xTower: A Multilingual LLM for Explaining and Correcting Translation Errors—0
Variable importance measure for spatial machine learning models with application to air pollution exposure prediction—0
Error Detection in Egocentric Procedural Task Videos—0
iSEA: An Interactive Pipeline for Semantic Error Analysis of NLP ModelsCode0
Score-CAM: Score-Weighted Visual Explanations for Convolutional Neural NetworksCode0
Understanding Humans' Strategies in Maze Solving—0
Show:102550
All datasetsCUB-200-2011CUB-200-2011 (ResNet-101)

Benchmark Results

#ModelMetricClaimedVerifiedStatus
1SMDL-Attribution (ICLR version)Average highest confidence (ResNet-101)0.45—Unverified
2Grad-CAM++Average highest confidence (ResNet-101)0.26—Unverified
3Score-CAMAverage highest confidence (ResNet-101)0.25—Unverified
4HSIC-AttributionAverage highest confidence (ResNet-101)0.25—Unverified
#ModelMetricClaimedVerifiedStatus
1Grad-CAM++Average highest confidence0.26—Unverified
2Score-CAMAverage highest confidence0.25—Unverified
3HSIC-AttributionAverage highest confidence0.25—Unverified