Ace.at_blacked.1.var Apr 2026

Deep Feature Consistent Variational Autoencoder - IEEE Xplore

: The variable represents a specific semantic direction that the ACE method attempts to remove or "erase" to prevent the model from generating undesirable images. ace.AT_Blacked.1.var

: The framework uses these features to improve the model's resistance to prompt-based attacks that try to bypass concept erasure. ace.AT_Blacked.1.var

: ACE introduces learnable gating mechanisms in the model's cross-attention layers, which are fine-tuned per concept using these deep feature representations. ace.AT_Blacked.1.var