Giovanni Bellitto

Instant, verifiable machine unlearning is now possible without retraining or access to original data, reshaping the landscape of privacy-compliant AI.

Forget gradient descent: this new method routes transformer activations through a Hopfield-inspired memory in a single forward pass to achieve state-of-the-art online continual learning.

Forget replaying old data – this continual learning method dreams up entirely new classes to train on, boosting forward transfer and outperforming standard rehearsal techniques.