Unlocking the Working Memory of Large Language Models for Latent Reasoningarxiv.org2 points·korbip··0 commentsOpen articleSaveView on HN