作者位次 #2/5ICML 2026Poster· Deep Learning / Large Language ModelsGrouter: Decoupling Routing from Representation for Accelerated MoE TrainingYuqi Xu · Rizhen Hu · zihan liu · Mou Sun · Kun Yuan可解释性效率与系统OpenReview
一作ICML 2026Poster· Deep Learning / Large Language ModelsSynergistic Intra- and Cross-Layer Regularization Losses for MoE Expert SpecializationRizhen Hu · Yuan Cao · Boao Kong · Mou Sun · Kun Yuan效率与系统OpenReview