The Continual Learning Problem

Great writeup. Are there any libraries that implement some of the methods described?

I appreciate that people are going beyond RAG and few shot prompting.

Rather than handcrafting solutions like it’s 1993, why not make robustness against forgetting part of the training objective?

Let the search algorithm figure it out.