This paper introduces SAFER, a continual unlearning framework designed to address critical issues like knowledge erosion and forgetting reversal in repeated machine unlearning scenarios. It aims to maintain model utility and prevent the reactivation of forgotten information, making AI systems more reliable and privacy-compliant over their lifecycle.

This paper introduces a novel approach for detecting synthetic fingerprints, framing it as a continual few-shot adaptation problem. It proposes using a combination of binary cross-entropy and supervised contrastive losses with experience replay to enable rapid adaptation to new synthetic data styles while mitigating catastrophic forgetting.

This paper introduces CogniSNN, a novel Spiking Neural Network (SNN) paradigm that incorporates Random Graph Architectures (RGA) to address the limitations of traditional, rigid SNN designs. CogniSNN enhances neuron-expandability, pathway-reusability, and dynamic-configurability, leading to improved performance, robustness, and continual learning capabilities in multi-task scenarios.