SafeCA reduces text-to-video jailbreak success by roughly 20% relative to T2VShield by masking anomalous cross-attention activations using clean-prompt statistics.
Transformer inter- pretability beyond attention visualization
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
SafeCA: Safe Cross-Attention Localization and Regulation for Text-to-Video Jailbreak Defense
SafeCA reduces text-to-video jailbreak success by roughly 20% relative to T2VShield by masking anomalous cross-attention activations using clean-prompt statistics.