CAPTAIN: Semantic Feature Injection for Memorization Mitigation in Text-to-Image Diffusion Models Paper • 2512.10655 • Published 20 days ago • 8
Shared Unsafe Directions Collection Do Language Models Share Unsafe Directions in Activation Space? • 5 items • Updated 15 days ago
Shared Unsafe Directions Collection Do Language Models Share Unsafe Directions in Activation Space? • 5 items • Updated 15 days ago
Shared Unsafe Directions Collection Do Language Models Share Unsafe Directions in Activation Space? • 5 items • Updated 15 days ago