Distillation Defenses Easily Break After Reinforcement Learning

(arxiv.org)

2 points | by ollybritton 11 hours ago ago