IC-834CLIP zero-shot predictions exhibit high equal opportunity difference when target and sensitive attributes are intrinsically dependent

Sepehr Dehdashtian, Lan Wang, Vishnu Boddeti

SourceFairerCLIP: Debiasing CLIP's Zero-Shot Predictions using Functions in RKHSs

On CelebA with high cheekbones as target and sex as sensitive attribute, zero-shot CLIP shows substantial bias measured by EOD. CLIP ResNet-50 achieves an EOD of 5.8% while CLIP ViT-L/14 achieves 2.8%, indicating that the model's predictions for cheekbones are systematically different for males versus females. This bias arises from a genuine correlation between the two attributes rather than a spurious one, and no existing debiasing baseline fully eliminates it.

Evidence
observational
Key metric
EOD: CLIP ResNet-50 5.8%, CLIP ViT-L/14 2.8% (CelebA, high cheekbones target, sex sensitive)
Model
CLIP / CLIP-ViT (LC)
Concepts
Failure mode
Datasets
CelebA [eval]
Methods
Equal Opportunity Difference [eval]
Related work
Contrastive adapter [compared-to], Linear Probing / Ridge regression linear probing / Linear probe / Linear probe fine-tuning / Linear regression probing / Linear ridge regression probes / Supervised probing / ERM linear probe [compared-to], DFR [compared-to]
Related findings
IC-835, IC-836, IC-837
Extraction
automatic-extraction