@inproceedings{luan-etal-2026-semfed,
title = "{S}em{F}ed: Semantic Federated Reinforcement Learning for Heterogeneous {LLM} Agents",
author = "Luan, Yuxuan and
Zhao, Zilin and
Zhang, Yihan and
Wang, Wei and
Sun, Chenyu and
Li, Lixiang and
Peng, Haipeng and
Zhang, Chaoning and
Xie, Lianku",
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.1147/",
pages = "24294--24308",
ISBN = "979-8-89176-462-0",
abstract = "Training Large Language Model (LLM) agents via Reinforcement Learning (RL) in federated settings faces fundamental challenges, including model heterogeneity across clients, prohibitive communication costs and severe data heterogeneity. We propose a semantic federated learning framework called SemFed that addresses all challenges by communicating behavioral patterns rather than model parameters. Each client extracts compact semantic policy summaries from local training trajectories, which are aggregated at the server into shared, elite and anti-patterns. These aggregated patterns are injected back into local training through different complementary channels. Since the communicated objects are structured behavioral abstractions, SemFed supports heterogeneous model architectures while reducing per-round communication to kilobytes. Experiments on ALFWorld and WebShop across multiple heterogeneous LLM backbones demonstrate that SemFed consistently surpasses strong federated baselines in task success while reducing communication overhead by orders of magnitude."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="luan-etal-2026-semfed">
<titleInfo>
<title>SemFed: Semantic Federated Reinforcement Learning for Heterogeneous LLM Agents</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yuxuan</namePart>
<namePart type="family">Luan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zilin</namePart>
<namePart type="family">Zhao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yihan</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Wei</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Chenyu</namePart>
<namePart type="family">Sun</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Lixiang</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Haipeng</namePart>
<namePart type="family">Peng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Chaoning</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Lianku</namePart>
<namePart type="family">Xie</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>Training Large Language Model (LLM) agents via Reinforcement Learning (RL) in federated settings faces fundamental challenges, including model heterogeneity across clients, prohibitive communication costs and severe data heterogeneity. We propose a semantic federated learning framework called SemFed that addresses all challenges by communicating behavioral patterns rather than model parameters. Each client extracts compact semantic policy summaries from local training trajectories, which are aggregated at the server into shared, elite and anti-patterns. These aggregated patterns are injected back into local training through different complementary channels. Since the communicated objects are structured behavioral abstractions, SemFed supports heterogeneous model architectures while reducing per-round communication to kilobytes. Experiments on ALFWorld and WebShop across multiple heterogeneous LLM backbones demonstrate that SemFed consistently surpasses strong federated baselines in task success while reducing communication overhead by orders of magnitude.</abstract>
<identifier type="citekey">luan-etal-2026-semfed</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.1147/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>24294</start>
<end>24308</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T SemFed: Semantic Federated Reinforcement Learning for Heterogeneous LLM Agents
%A Luan, Yuxuan
%A Zhao, Zilin
%A Zhang, Yihan
%A Wang, Wei
%A Sun, Chenyu
%A Li, Lixiang
%A Peng, Haipeng
%A Zhang, Chaoning
%A Xie, Lianku
%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 luan-etal-2026-semfed
%X Training Large Language Model (LLM) agents via Reinforcement Learning (RL) in federated settings faces fundamental challenges, including model heterogeneity across clients, prohibitive communication costs and severe data heterogeneity. We propose a semantic federated learning framework called SemFed that addresses all challenges by communicating behavioral patterns rather than model parameters. Each client extracts compact semantic policy summaries from local training trajectories, which are aggregated at the server into shared, elite and anti-patterns. These aggregated patterns are injected back into local training through different complementary channels. Since the communicated objects are structured behavioral abstractions, SemFed supports heterogeneous model architectures while reducing per-round communication to kilobytes. Experiments on ALFWorld and WebShop across multiple heterogeneous LLM backbones demonstrate that SemFed consistently surpasses strong federated baselines in task success while reducing communication overhead by orders of magnitude.
%U https://aclanthology.org/2026.findings-emnlp.1147/
%P 24294-24308
Markdown (Informal)
[SemFed: Semantic Federated Reinforcement Learning for Heterogeneous LLM Agents](https://aclanthology.org/2026.findings-emnlp.1147/) (Luan et al., Findings 2026)
ACL
- Yuxuan Luan, Zilin Zhao, Yihan Zhang, Wei Wang, Chenyu Sun, Lixiang Li, Haipeng Peng, Chaoning Zhang, and Lianku Xie. 2026. SemFed: Semantic Federated Reinforcement Learning for Heterogeneous LLM Agents. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 24294–24308, Budapest, Hungary. Association for Computational Linguistics.