English(EN) Experts Exhibit Structured Redundancy: Structure-Preserving Expert Pruning for Mixture-of-Experts Large Language Models

新的剪枝方法在保持性能的同时减小了MoE大型语言模型的内存占用

作者 PulseAugur 编辑部 · [1 个来源] · 2026-11-02 05:00

研究人员开发了用于混合专家（MoE）大型语言模型中剪枝专家的新方法，解决了存储所有专家带来的显著内存开销问题。提出的技术，即投影能量专家剪枝（PEEP）和分布覆盖专家剪枝（DCEP），旨在保留专家输出的结构和功能子空间。在DeepSeek-V2-Lite和Qwen3-30B-A3B模型上的实验表明，即使只保留50%或75%的专家，这些方法在各种基准测试中保持性能方面也优于现有基线。 AI

影响这些剪枝技术可以显著降低部署大型语言模型的计算和内存成本，使其更易于访问。

排序理由该集群包含一篇详细介绍优化大型语言模型新方法的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.AI 阅读 →

AI 生成摘要 · Google Gemini · 来自 1 个来源。我们如何撰写摘要 →

本文如何被排名

Signal score: 1 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 该集群包含一篇详细介绍优化大型语言模型新方法的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration: Single-source cluster; Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics: paper, infra; Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance: High; Clearly on-topic for AI-industry coverage.
Story freshness: 1 days old; Coverage has settled into its steady-state source set.

完整方法见我们的编辑标准。

报道来源 [1]

arXiv cs.AI TIER_1 English(EN) · Tian Wang, Yongding Tao, Yihong Dong, Xiaoming Wu, Ying Guo, Ge Li · 2026-11-02 05:00

Experts Exhibit Structured Redundancy: Structure-Preserving Expert Pruning for Mixture-of-Experts Large Language Models

arXiv:2610.41236v1 Announce Type: new Abstract: Sparse MoE mechanism significantly improves the trade-off between model capacity and inference efficiency. However, practical deployment still requires storing all experts, resulting in substantial memory overhead. We show that expe…

本文如何被排名

报道来源 [1]

Experts Exhibit Structured Redundancy: Structure-Preserving Expert Pruning for Mixture-of-Experts Large Language Models

相关实体

相关话题