List your company
ResearchFormerly WizWand

Spurious Correlation Discovery on MNIST, FashionMNIST, and COCO 654 evaluation settings (test)

91.1mP@10RAVL
-2.08422.10846.370.492Nov 6, 2024
Updated 8mo ago

Evaluation results

MethodLinks
RAVLCorrelation Strength (τ_eval)=40, Num. Eval Settings=1682024.1191.1
RAVLCorrelation Strength (τ_eval)=30, Num. Eval Settings=2342024.1184.2
RAVLCorrelation Strength (τ_eval)=20, Num. Eval Settings=3692024.1176.2
RAVLCorrelation Strength (τ_eval)=10, Num. Eval Settings=6542024.1161.8
Spurious-Aware DetectionCorrelation Strength (τ_eval)=40, Num. Eval Settings=1682024.1142
Spurious-Aware DetectionCorrelation Strength (τ_eval)=30, Num. Eval Settings=2342024.1132.1
Spurious-Aware DetectionCorrelation Strength (τ_eval)=20, Num. Eval Settings=3692024.1125.3
RandomCorrelation Strength (τ_eval)=10, Num. Eval Settings=6542024.1121.2
Distilling FailuresCorrelation Strength (τ_eval)=10, Num. Eval Settings=6542024.1120.1
Spurious-Aware DetectionCorrelation Strength (τ_eval)=10, Num. Eval Settings=6542024.1120
GeorgeCorrelation Strength (τ_eval)=10, Num. Eval Settings=6542024.1119.3
RandomCorrelation Strength (τ_eval)=20, Num. Eval Settings=3692024.1118.2
DominoCorrelation Strength (τ_eval)=10, Num. Eval Settings=6542024.1117.1
Distilling FailuresCorrelation Strength (τ_eval)=20, Num. Eval Settings=3692024.1116.2
GeorgeCorrelation Strength (τ_eval)=20, Num. Eval Settings=3692024.1115.9
RandomCorrelation Strength (τ_eval)=30, Num. Eval Settings=2342024.1115.5
DominoCorrelation Strength (τ_eval)=20, Num. Eval Settings=3692024.1115
RandomCorrelation Strength (τ_eval)=40, Num. Eval Settings=1682024.1112.5
DominoCorrelation Strength (τ_eval)=30, Num. Eval Settings=2342024.1111.7
GeorgeCorrelation Strength (τ_eval)=30, Num. Eval Settings=2342024.1110.9
DominoCorrelation Strength (τ_eval)=40, Num. Eval Settings=1682024.119
Distilling FailuresCorrelation Strength (τ_eval)=30, Num. Eval Settings=2342024.118.5
GeorgeCorrelation Strength (τ_eval)=40, Num. Eval Settings=1682024.117.7
Distilling FailuresCorrelation Strength (τ_eval)=40, Num. Eval Settings=1682024.111.5