First accepted paper of the year: "From Attention to Activation: Unraveling the Enigmas of Large Language Models" has been accepted to ICLR 2025. The most educative paper I have co-wrote, it strengthens some claims known in the community, it opposes others, 1/n
