AI RESEARCH

Safety-Potential Pruning for Enhancing Safety Prompts Against VLM Jailbreaking Without Retraining

arXiv CS.CV

ArXi:2603.14219v1 Announce Type: new Safety prompts constitute an interpretable layer of defense against jailbreak attacks in vision-language models (VLMs); however, their efficacy is constrained by the models' latent structural responsiveness. We observe that such prompts consistently engage a sparse set of parameters that remain largely quiescent during benign use. This finding motivates the Safety Subnetwork Hypothesis: VLMs embed structurally distinct pathways capable of enforcing safety, but these pathways remain dormant without explicit stimulation.