Powered by OpenAIRE graph
Found an issue? Give us feedback
image/svg+xml art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos Open Access logo, converted into svg, designed by PLoS. This version with transparent background. http://commons.wikimedia.org/wiki/File:Open_Access_logo_PLoS_white.svg art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos http://www.plos.org/ ZENODOarrow_drop_down
image/svg+xml art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos Open Access logo, converted into svg, designed by PLoS. This version with transparent background. http://commons.wikimedia.org/wiki/File:Open_Access_logo_PLoS_white.svg art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos http://www.plos.org/
ZENODO
Software
Data sources: ZENODO
addClaim

CGAA: Concept-Guided Adversarial Attacks

Authors: Mukhtar, Muhammad Taha;

CGAA: Concept-Guided Adversarial Attacks

Abstract

An adversarial attack that adds a CAV-based concept-alignment term to the BIM loss. Archived after Phase 1 smoke tests revealed that the approach moves linear probe outputs without producing genuine concept changes (mean-pool gradient pathology), and that the core entanglement claim was occupied by Nicolson et al. (TMLR 2025).

Powered by OpenAIRE graph
Found an issue? Give us feedback