HEAPr: Hessian-based Efficient Atomic Expert Pruning in Output Space 事件

HEAPr: Hessian-based Efficient Atomic Expert Pruning in Output Space arXiv:2509.22299v3 Announce Type: replace-cross Abstract: Mixture-of-Experts (MoE) architectures in large language models (LLMs) deliver exceptional performance and reduced inference costs compared to dense LLMs. However, their large parameter counts result in prohibitive memory requirements, limiting practical deployment. While existing pruning methods primarily focus on expert-level pruning, this coarse granularity often lea