@inproceedings{do-etal-2025-simsmoe,
title = "{S}im{SM}o{E}: Toward Efficient Training Mixture of Experts via Solving Representational Collapse",
author = "Do, Giang and
Le, Hung and
Tran, Truyen",
editor = "Chiruzzo, Luis and
Ritter, Alan and
Wang, Lu",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2025",
month = apr,
year = "2025",
address = "Albuquerque, New Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.findings-naacl.107/",
doi = "10.18653/v1/2025.findings-naacl.107",
pages = "2012--2025",
ISBN = "979-8-89176-195-7",
abstract = "Sparse mixture of experts (SMoE) have emerged as an effective approach for scaling large language models while keeping a constant computational cost. Regardless of several notable successes of SMoE, effective training such architecture remains elusive due to the representation collapse problem, which in turn harms model performance and causes parameter redundancy. In this work, we present Similarity-based Sparse Mixture of Experts (SimSMoE), a novel similarity of neural network algorithm, that guarantees a solution to address the representation collapse issue between experts given a fixed FLOPs budget. We conduct extensive empirical evaluations on three large language models for both Pre-training and Fine-tuning tasks to illustrate the efficacy, robustness, and scalability of our method. The results demonstrate that SimSMoE significantly enhances existing routing policy and outperforms other SMoE routing methods in performance for the tasks. Our implementation is publicly available at https://github.com/giangdip2410/SimSMoE."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="do-etal-2025-simsmoe">
<titleInfo>
<title>SimSMoE: Toward Efficient Training Mixture of Experts via Solving Representational Collapse</title>
</titleInfo>
<name type="personal">
<namePart type="given">Giang</namePart>
<namePart type="family">Do</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hung</namePart>
<namePart type="family">Le</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Truyen</namePart>
<namePart type="family">Tran</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-04</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: NAACL 2025</title>
</titleInfo>
<name type="personal">
<namePart type="given">Luis</namePart>
<namePart type="family">Chiruzzo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Alan</namePart>
<namePart type="family">Ritter</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Lu</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Albuquerque, New Mexico</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-195-7</identifier>
</relatedItem>
<abstract>Sparse mixture of experts (SMoE) have emerged as an effective approach for scaling large language models while keeping a constant computational cost. Regardless of several notable successes of SMoE, effective training such architecture remains elusive due to the representation collapse problem, which in turn harms model performance and causes parameter redundancy. In this work, we present Similarity-based Sparse Mixture of Experts (SimSMoE), a novel similarity of neural network algorithm, that guarantees a solution to address the representation collapse issue between experts given a fixed FLOPs budget. We conduct extensive empirical evaluations on three large language models for both Pre-training and Fine-tuning tasks to illustrate the efficacy, robustness, and scalability of our method. The results demonstrate that SimSMoE significantly enhances existing routing policy and outperforms other SMoE routing methods in performance for the tasks. Our implementation is publicly available at https://github.com/giangdip2410/SimSMoE.</abstract>
<identifier type="citekey">do-etal-2025-simsmoe</identifier>
<identifier type="doi">10.18653/v1/2025.findings-naacl.107</identifier>
<location>
<url>https://aclanthology.org/2025.findings-naacl.107/</url>
</location>
<part>
<date>2025-04</date>
<extent unit="page">
<start>2012</start>
<end>2025</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T SimSMoE: Toward Efficient Training Mixture of Experts via Solving Representational Collapse
%A Do, Giang
%A Le, Hung
%A Tran, Truyen
%Y Chiruzzo, Luis
%Y Ritter, Alan
%Y Wang, Lu
%S Findings of the Association for Computational Linguistics: NAACL 2025
%D 2025
%8 April
%I Association for Computational Linguistics
%C Albuquerque, New Mexico
%@ 979-8-89176-195-7
%F do-etal-2025-simsmoe
%X Sparse mixture of experts (SMoE) have emerged as an effective approach for scaling large language models while keeping a constant computational cost. Regardless of several notable successes of SMoE, effective training such architecture remains elusive due to the representation collapse problem, which in turn harms model performance and causes parameter redundancy. In this work, we present Similarity-based Sparse Mixture of Experts (SimSMoE), a novel similarity of neural network algorithm, that guarantees a solution to address the representation collapse issue between experts given a fixed FLOPs budget. We conduct extensive empirical evaluations on three large language models for both Pre-training and Fine-tuning tasks to illustrate the efficacy, robustness, and scalability of our method. The results demonstrate that SimSMoE significantly enhances existing routing policy and outperforms other SMoE routing methods in performance for the tasks. Our implementation is publicly available at https://github.com/giangdip2410/SimSMoE.
%R 10.18653/v1/2025.findings-naacl.107
%U https://aclanthology.org/2025.findings-naacl.107/
%U https://doi.org/10.18653/v1/2025.findings-naacl.107
%P 2012-2025
Markdown (Informal)
[SimSMoE: Toward Efficient Training Mixture of Experts via Solving Representational Collapse](https://aclanthology.org/2025.findings-naacl.107/) (Do et al., Findings 2025)
ACL