Proto-Clip

Repurposing Adversarial Perturbations for Continual Learning: From Defense to Active Alignment

Announce Type: new Abstract: In dynamic environments, large language models need to keep adapting to new tasks, but continual learning often suffers from forgetting, limited transfer, and vulnerability to adversarial perturbations. To address this, we present AdvCL, which repurposes adversarial perturbations as a geometric control signal for stable continual adaptation.