Fetching the paper…

ReLU$^2$ Wins: Discovering Efficient Activation Functions for Sparse LLMs · Around