CAN LLMS LEARN FROM A SINGLE EXAMPLE?

Summary from the TLDR AI newsletter

It appears that AI models were able to rapidly memorize examples from the dataset after seeing them just once. This astonishing feat contradicts most prior wisdom about neural network sample efficiency.