#personalization

Verifiable model unlearning on edge devices without retraining

ZK-APEX combines sparse masking and zero-knowledge proofs to let providers verify that personalized models forget targeted data while preserving local utility.