作者位次 #2/4ICML 2026Poster· Deep Learning / Attention Mechanisms
Attention Sink Forges Native MoE in Attention Layers: Sink-Aware Training to Address Head Collapse
Zizhuo Fu · Wenxuan Zeng · Runsheng Wang · Meng Li
首次获取该候选人的完整档案及 StartLux 评估共 20 积分;每位用户单独解锁,之后查看免费。主动刷新 20 积分。
实时查找电子邮箱,费用不超过20积分;未找到邮箱不扣费。
Zizhuo Fu · Wenxuan Zeng · Runsheng Wang · Meng Li
Jiangrui Yu · Ye Yu · Si Chen · Chenqi Lin · Wenxuan Zeng · Junfeng Fan · Mingyu Gao · Meng Li