Modelpedia
Work in progress
About
Findings
Models
Concepts
Methods
Datasets
Sources
Light
Dark
Orthogonal Matching Pursuit
anchor
Findings
IC-151
Each CLIP neuron's second-order effect is approximately a single linear direction in the joint text-image space, significant for fewer than 2% of images
[primary]
IC-152
CLIP's polysemantic neurons encode spurious correlations between unrelated concepts that can be exploited to generate adversarial misclassifications
[primary]