Concetto Spampinato

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.

You can now audit black-box vision models for biases and failure modes using only their output probabilities, thanks to a clever LLM-powered semantic search.

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