shadow-alignment

Tag

Cards List
#shadow-alignment

Decoupled Alignment for Robust Plug-and-Play Adaptation

arXiv cs.CL · yesterday Cached

Introduces a training-free method for enhancing safety alignment of LLMs by using knowledge distillation and model fusion to prevent shadow alignment, improving defense success rate by 14.42% on harmful question datasets without compromising performance.

0 favorites 0 likes
← Back to home

Submit Feedback