<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="park-etal-2026-boundary">
    <titleInfo>
        <title>Boundary-Aware Expert Ordering Preservation with Next-Router Supervision for MoE Quantization</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Hancheol</namePart>
        <namePart type="family">Park</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Geonho</namePart>
        <namePart type="family">Lee</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Tae-Ho</namePart>
        <namePart type="family">Kim</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <originInfo>
        <dateIssued>2026-10</dateIssued>
    </originInfo>
    <typeOfResource>text</typeOfResource>
    <relatedItem type="host">
        <titleInfo>
            <title>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Hung-Yi</namePart>
            <namePart type="family">Lee</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Sunipa</namePart>
            <namePart type="family">Dev</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Jessy</namePart>
            <namePart type="family">Li</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Desmond</namePart>
            <namePart type="family">Elliott</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <originInfo>
            <publisher>Association for Computational Linguistics</publisher>
            <place>
                <placeTerm type="text">Budapest, Hungary</placeTerm>
            </place>
        </originInfo>
        <genre authority="marcgt">conference publication</genre>
        <identifier type="isbn">979-8-89176-459-0</identifier>
    </relatedItem>
    <abstract>Mixture-of-Experts (MoE) language models scale model capacity through sparse expert activation, but their large parameter footprint remains costly to deploy. Post-training quantization (PTQ) offers a practical compression strategy; however, quantization errors can alter router decisions and cause the quantized model to activate experts different from those selected by its full-precision counterpart. Prior router-aware PTQ objectives mitigate such expert-shift by aligning router outputs or preserving top-k expert orderings and margins at a given router. Yet sequential block-wise PTQ introduces an additional failure mode: quantizing a decoder block changes the representations received by the next MoE router, potentially causing downstream expert-shift that is not directly optimized by existing objectives. To address this issue, we propose MENDS-MoE (Margin-Preserving Expert Orderings with Next-Router Decision Supervision), an auxiliary calibration objective that supervises current-block quantization using routing decisions at the next MoE router. MENDS-MoE further introduces a boundary-aware margin loss that preserves orderings among selected experts while maintaining their separation from unselected experts near the top-k boundary. It is applied only during calibration and adds no inference-time computation. Across three MoE language models under 4-bit and 3-bit-average weight-only quantization, MENDS-MoE achieves the best average downstream accuracy and the lowest perplexity in most evaluated model–bit settings. Ablation studies further support the contributions of both next-router supervision and boundary-aware margin preservation.</abstract>
    <identifier type="citekey">park-etal-2026-boundary</identifier>
    <location>
        <url>https://aclanthology.org/2026.emnlp-main.2041/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>46537</start>
            <end>46548</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T Boundary-Aware Expert Ordering Preservation with Next-Router Supervision for MoE Quantization
%A Park, Hancheol
%A Lee, Geonho
%A Kim, Tae-Ho
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-459-0
%F park-etal-2026-boundary
%X Mixture-of-Experts (MoE) language models scale model capacity through sparse expert activation, but their large parameter footprint remains costly to deploy. Post-training quantization (PTQ) offers a practical compression strategy; however, quantization errors can alter router decisions and cause the quantized model to activate experts different from those selected by its full-precision counterpart. Prior router-aware PTQ objectives mitigate such expert-shift by aligning router outputs or preserving top-k expert orderings and margins at a given router. Yet sequential block-wise PTQ introduces an additional failure mode: quantizing a decoder block changes the representations received by the next MoE router, potentially causing downstream expert-shift that is not directly optimized by existing objectives. To address this issue, we propose MENDS-MoE (Margin-Preserving Expert Orderings with Next-Router Decision Supervision), an auxiliary calibration objective that supervises current-block quantization using routing decisions at the next MoE router. MENDS-MoE further introduces a boundary-aware margin loss that preserves orderings among selected experts while maintaining their separation from unselected experts near the top-k boundary. It is applied only during calibration and adds no inference-time computation. Across three MoE language models under 4-bit and 3-bit-average weight-only quantization, MENDS-MoE achieves the best average downstream accuracy and the lowest perplexity in most evaluated model–bit settings. Ablation studies further support the contributions of both next-router supervision and boundary-aware margin preservation.
%U https://aclanthology.org/2026.emnlp-main.2041/
%P 46537-46548

Markdown (Informal)

[Boundary-Aware Expert Ordering Preservation with Next-Router Supervision for MoE Quantization](https://aclanthology.org/2026.emnlp-main.2041/) (Park et al., EMNLP 2026)

• Boundary-Aware Expert Ordering Preservation with Next-Router Supervision for MoE Quantization (Park et al., EMNLP 2026)

ACL

• Hancheol Park, Geonho Lee, and Tae-Ho Kim. 2026. Boundary-Aware Expert Ordering Preservation with Next-Router Supervision for MoE Quantization. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 46537–46548, Budapest, Hungary. Association for Computational Linguistics.