claim
active
claim:adversarial-attacks-work-by-activating-many-interfering-features-in-superposition-adversarial-training-reorganizes-geometry-to-prevent-this

Adversarial attacks work by activating many interfering features in superposition; adversarial training reorganizes geometry to prevent this.

Interpretation of Figure 3.

Source paper

extracted_from
Adversarial Examples Are Not Bugs, They Are Superposition
(2025) · Gorton, Liv · Lewis, Owen

Neighborhood — ranked by edge-count