List your company
ResearchFormerly WizWand

Image Classification on Waterbirds 100% correlation (test)

91.9Worst-group AccuracyBAP (LP)
15.77235.53655.375.064Aug 12, 2025Sep 26, 2025Nov 10, 2025Dec 26, 2025Feb 9, 2026Mar 26, 2026May 11, 2026
Updated 5mo ago

Evaluation results

MethodLinks
BAP (LP)Backbone=ConvNeXT-W, Pre-training=CLIP, Evaluation Protocol=Linear Probing2026.0591.994
Data-Matched ControlBackbone=ConvNeXT-W, Pre-training=CLIP, Evaluation Protocol=Trained Evaluation2026.0580.787.2
SupERGroup Info=false, Train Twice=false2025.0879.785
BAP (ZS)Backbone=ConvNeXT-W, Pre-training=CLIP, Evaluation Protocol=Zero-Shot2026.0571.686.4
DIALBackbone=ConvNeXT-W, Pre-training=CLIP, Evaluation Protocol=Zero-Shot2026.0564.979.7
CnCGroup Info=false, Train Twice=true2025.0862.181.9
LfFGroup Info=false, Train Twice=false2025.0861.580.6
JTTGroup Info=false, Train Twice=true2025.0861.379.7
RoboShotBackbone=ConvNeXT-W, Pre-training=CLIP, Evaluation Protocol=Zero-Shot2026.0559.275.8
CVaR DROGroup Info=false, Train Twice=false2025.085879
IRMGroup Info=true, Train Twice=false2025.085780.5
GroupDROGroup Info=true, Train Twice=false2025.0856.579.4
UWGroup Info=true, Train Twice=false2025.0856.478.6
GALSGroup Info=false, Train Twice=false2025.085579.7
DFRGroup Info=true, Train Twice=true2025.0848.276.4
Native BackboneBackbone=ConvNeXT-W, Pre-training=CLIP, Evaluation Protocol=Zero-Shot2026.0547.275.3
ERMGroup Info=false, Train Twice=false2025.0846.474.8
PruSCBackbone=ConvNeXT-W, Pre-training=CLIP, Evaluation Protocol=Trained Evaluation2026.0535.647.3
Native Backbone (LP)Backbone=ConvNeXT-W, Pre-training=CLIP, Evaluation Protocol=Linear Probing2026.0531.363.3
WiSE-FTBackbone=ConvNeXT-W, Pre-training=CLIP, Evaluation Protocol=Trained Evaluation2026.0521.663.1
ERM (LP-FT)Backbone=ConvNeXT-W, Pre-training=CLIP, Evaluation Protocol=Linear Probing & Fine-Tuning2026.0518.763.1