paper-with-me

홈 › Papers

On DeepSeekMoE: Statistical Benefits of Shared Experts and Normalized Sigmoid Gating

2025-05-16 · Huy Nguyen, Thong T. Doan, Quang Pham, Nghi D. Q. Bui, Nhat Ho, Alessandro Rinaldo

Mixture of experts (MoE) methods are a key component in most large language model architectures, including the recent series of DeepSeek models. Compared to other MoE implementations, DeepSeekMoE stands out because of two unique features: the deployment of a shared expert strategy and of the normalized sigmoid gating mechanism. Despite the prominent role of DeepSeekMoE in the success of the DeepSeek series of models, there have been only a few attempts to justify theoretically the value of the shared expert strategy, while its normalized sigmoid gating has remained unexplored. To bridge this gap, we undertake a comprehensive theoretical study of these two features of DeepSeekMoE from a statistical perspective. We perform a convergence analysis of the expert estimation task to highlight the gains in sample efficiency for both the shared expert strategy and the normalized sigmoid gating, offering useful insights into the design of expert and gating structures. To verify empirically our theoretical findings, we carry out several experiments on both synthetic data and real-world datasets for (vision) language modeling tasks. Finally, we conduct an extensive empirical analysis of the router behaviors, ranging from router saturation, router change rate, to expert utilization.

📄 PDF Abstract BibTeX arXiv:2505.10860

Code (0)

등록된 구현이 없습니다.

Tasks

Language ModelingLanguage ModellingLarge Language ModelMixture-of-Experts

Methods 이 논문이 사용한 방법론

MoE 설명 없음

Similar Papers 제목 키워드 기반

DeepSeekMoE: Towards Ultimate Expert Specialization in Mixture-of-Experts Language Models

2024-01-11 · Damai Dai, Chengqi Deng, Chenggang Zhao, R. X. Xu 외

In the era of large language models, Mixture-of-Experts (MoE) is a promising architecture for managing computational costs when scaling up model parameters. However, conventional MoE architectures like GShard, which acti…

Language ModellingLarge Language ModelMixture-of-Experts

Condense, Don't Just Prune: Enhancing Efficiency and Performance in MoE Layer Pruning

2024-11-26 · Mingyu Cao, Gen Li, Jie Ji, JiaQi Zhang 외

Mixture-of-Experts (MOE) has garnered significant attention for their ability to scale up neural networks while utilizing the same or even fewer active parameters. However, MoE does not relieve the massive memory require…

Mixture-of-Experts

Synergistic Intra- and Cross-Layer Regularization Losses for MoE Expert Specialization

2026-02-15 · Rizhen Hu, Yuan Cao, Boao Kong, Mou Sun 외 arxiv

Sparse Mixture-of-Experts (MoE) models scale Transformers efficiently but suffer from expert overlap -- redundant representations across experts and routing ambiguity, resulting in severely underutilized model capacity. …

HELLoRA: Hot Experts Layer-Level Low-Rank Adaptation for Mixture-of-Experts Models

2026-05-11 · Jia Wei, Zhonghao Zhang, Ping Chen, Qianyang li 외 arxiv

Low-Rank Adaptation (LoRA) dominates parameter-efficient fine-tuning of large language models, yet most variants target dense architectures. Mixture-of-Experts (MoE) models scale parameters at near-constant per-token com…

parameter-efficient fine-tuningMathematical ReasoningCode Generation

DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Model

2024-05-07 · DeepSeek-AI, Aixin Liu, Bei Feng, Bin Wang 외

We present DeepSeek-V2, a strong Mixture-of-Experts (MoE) language model characterized by economical training and efficient inference. It comprises 236B total parameters, of which 21B are activated for each token, and su…

Language ModelingLanguage ModellingMixture-of-ExpertsReinforcement Learning (RL)