🚨 Your diffusion model might relearn what you made it forget, even without trying! 🚨 In new work, we show concept unlearning can be reversed when finetuning on unrelated data. Even worse, unlearned concepts can resurface when prompting for an unrelated concept! (1/9)

New preprint with Joao Cavalcanti and @Ashia__Wilson! arxiv.org/abs/2408.13150 tl;dr: If you make your backtracking line search routine ADAPTIVE, you get faster continuous optimization algorithms with no additional computational burden! Longer summary below!🧵