@inproceedings{rongfeng-etal-2026-meta,
title = "Meta-Learning Where to Allocate Experts: Task-Conditioned Layer-Wise Compression for {M}o{E}s",
author = "Rongfeng, Wang and
Weng, Shichao and
Wang, Zhiqiang and
Liu, Xinyu and
Yi, Yang and
Zhou, Peilong and
Hongwei, Tang",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {EMNLP} 2026",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-emnlp.1220/",
pages = "25830--25842",
ISBN = "979-8-89176-462-0",
abstract = "Mixture-of-Experts (MoE) models route each token to a subset of expert networks, increasing capacity while keeping per-token computation sparse. Many deployed MoEs use a fixed number of active experts across layers and tasks, despite evidence that layer roles, expert redundancy, and required capacity vary with depth and input difficulty. Existing layer-wise methods usually determine allocations offline and reuse them across tasks, while token-level methods rely on local routing signals without task-level context. We propose MetaNet, a support-set controller that predicts an expert-retention threshold and a bounded routing bias for each layer while keeping the backbone, experts, and original router frozen. On DeepSeek-MoE-16B-Chat, MetaNet provides a tunable accuracy{--}expert-activation trade-off. Relative to fixed k=6, a conservative setting activates 3.61 experts on average (40{\%} fewer) with comparable MMLU accuracy (0.489 versus 0.474); an aggressive setting activates 2.28 experts on average (62{\%} fewer) at 0.438 accuracy. Trained on MMLU, MetaNet transfers to C-Eval without retraining, activating 2.90 experts on average (52{\%} fewer than fixed k=6) at 0.386 accuracy."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="rongfeng-etal-2026-meta">
<titleInfo>
<title>Meta-Learning Where to Allocate Experts: Task-Conditioned Layer-Wise Compression for MoEs</title>
</titleInfo>
<name type="personal">
<namePart type="given">Wang</namePart>
<namePart type="family">Rongfeng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shichao</namePart>
<namePart type="family">Weng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhiqiang</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xinyu</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yang</namePart>
<namePart type="family">Yi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Peilong</namePart>
<namePart type="family">Zhou</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tang</namePart>
<namePart type="family">Hongwei</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>Findings of the Association for Computational Linguistics: EMNLP 2026</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">979-8-89176-462-0</identifier>
</relatedItem>
<abstract>Mixture-of-Experts (MoE) models route each token to a subset of expert networks, increasing capacity while keeping per-token computation sparse. Many deployed MoEs use a fixed number of active experts across layers and tasks, despite evidence that layer roles, expert redundancy, and required capacity vary with depth and input difficulty. Existing layer-wise methods usually determine allocations offline and reuse them across tasks, while token-level methods rely on local routing signals without task-level context. We propose MetaNet, a support-set controller that predicts an expert-retention threshold and a bounded routing bias for each layer while keeping the backbone, experts, and original router frozen. On DeepSeek-MoE-16B-Chat, MetaNet provides a tunable accuracy–expert-activation trade-off. Relative to fixed k=6, a conservative setting activates 3.61 experts on average (40% fewer) with comparable MMLU accuracy (0.489 versus 0.474); an aggressive setting activates 2.28 experts on average (62% fewer) at 0.438 accuracy. Trained on MMLU, MetaNet transfers to C-Eval without retraining, activating 2.90 experts on average (52% fewer than fixed k=6) at 0.386 accuracy.</abstract>
<identifier type="citekey">rongfeng-etal-2026-meta</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.1220/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>25830</start>
<end>25842</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Meta-Learning Where to Allocate Experts: Task-Conditioned Layer-Wise Compression for MoEs
%A Rongfeng, Wang
%A Weng, Shichao
%A Wang, Zhiqiang
%A Liu, Xinyu
%A Yi, Yang
%A Zhou, Peilong
%A Hongwei, Tang
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Findings of the Association for Computational Linguistics: EMNLP 2026
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-462-0
%F rongfeng-etal-2026-meta
%X Mixture-of-Experts (MoE) models route each token to a subset of expert networks, increasing capacity while keeping per-token computation sparse. Many deployed MoEs use a fixed number of active experts across layers and tasks, despite evidence that layer roles, expert redundancy, and required capacity vary with depth and input difficulty. Existing layer-wise methods usually determine allocations offline and reuse them across tasks, while token-level methods rely on local routing signals without task-level context. We propose MetaNet, a support-set controller that predicts an expert-retention threshold and a bounded routing bias for each layer while keeping the backbone, experts, and original router frozen. On DeepSeek-MoE-16B-Chat, MetaNet provides a tunable accuracy–expert-activation trade-off. Relative to fixed k=6, a conservative setting activates 3.61 experts on average (40% fewer) with comparable MMLU accuracy (0.489 versus 0.474); an aggressive setting activates 2.28 experts on average (62% fewer) at 0.438 accuracy. Trained on MMLU, MetaNet transfers to C-Eval without retraining, activating 2.90 experts on average (52% fewer than fixed k=6) at 0.386 accuracy.
%U https://aclanthology.org/2026.findings-emnlp.1220/
%P 25830-25842
Markdown (Informal)
[Meta-Learning Where to Allocate Experts: Task-Conditioned Layer-Wise Compression for MoEs](https://aclanthology.org/2026.findings-emnlp.1220/) (Rongfeng et al., Findings 2026)
ACL