@inproceedings{zheng-etal-2026-routepo,
title = "{R}oute{PO}: Personalized Routing Policy Optimization for {LLM}-based Multi-Channel Recall",
author = "Zheng, Zaiyi and
Zhu, Yaochen and
Ju, Mingxuan and
Zhao, Tong and
Shah, Neil and
Li, Jundong",
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.784/",
pages = "16593--16609",
ISBN = "979-8-89176-462-0",
abstract = "LLM-based rerankers in recommender systems are inherently constrained by limited context windows: they can only reorder a small retrieved candidate set, so any relevant item missed by upstream retrieval cannot be recovered at ranking time. To mitigate this dependency, we propose RoutePO, which repositions the LLM from an item ranker to a recall strategy router for multi-channel recall. Conditioned on a user{'}s profile/context (interaction history and optional demographics), the LLM outputs personalized fusion weights over heterogeneous retrieval experts, enabling dynamic candidate aggregation without exposing long candidate lists from the entire item pool to the prompt. We train the router via a synergistic two-stage alignment process: Coarse Routing Alignment (SFT) warm-starts the policy with a quality-aware oracle prior over retrieval experts, and Fine-grained Metric Optimization (RL) directly optimizes recall-stage ranking metrics (e.g., NDCG) by sampling simplex-valued routing weights and rewarding fused-list quality. To support continuous exploration over the simplex under GRPO-style optimization, we introduce SofT-GRPO based on Gumbel{--}Softmax reparameterization with nucleus restriction and KL regularization. Finally, we instantiate fusion with both weighted score merging and a weight-aware interleaving strategy (Snake Fusion) that preserves within-channel rankings. Experiments on three real-world datasets demonstrate consistent gains over strong individual recallers and multi-channel fusion baselines, including up to 15.88{\%} relative improvement in NDCG@50 on Food. Our implementation is available on \url{https://github.com/zhengzaiyi/RoutePO}."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="zheng-etal-2026-routepo">
<titleInfo>
<title>RoutePO: Personalized Routing Policy Optimization for LLM-based Multi-Channel Recall</title>
</titleInfo>
<name type="personal">
<namePart type="given">Zaiyi</namePart>
<namePart type="family">Zheng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yaochen</namePart>
<namePart type="family">Zhu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mingxuan</namePart>
<namePart type="family">Ju</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tong</namePart>
<namePart type="family">Zhao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Neil</namePart>
<namePart type="family">Shah</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jundong</namePart>
<namePart type="family">Li</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>LLM-based rerankers in recommender systems are inherently constrained by limited context windows: they can only reorder a small retrieved candidate set, so any relevant item missed by upstream retrieval cannot be recovered at ranking time. To mitigate this dependency, we propose RoutePO, which repositions the LLM from an item ranker to a recall strategy router for multi-channel recall. Conditioned on a user’s profile/context (interaction history and optional demographics), the LLM outputs personalized fusion weights over heterogeneous retrieval experts, enabling dynamic candidate aggregation without exposing long candidate lists from the entire item pool to the prompt. We train the router via a synergistic two-stage alignment process: Coarse Routing Alignment (SFT) warm-starts the policy with a quality-aware oracle prior over retrieval experts, and Fine-grained Metric Optimization (RL) directly optimizes recall-stage ranking metrics (e.g., NDCG) by sampling simplex-valued routing weights and rewarding fused-list quality. To support continuous exploration over the simplex under GRPO-style optimization, we introduce SofT-GRPO based on Gumbel–Softmax reparameterization with nucleus restriction and KL regularization. Finally, we instantiate fusion with both weighted score merging and a weight-aware interleaving strategy (Snake Fusion) that preserves within-channel rankings. Experiments on three real-world datasets demonstrate consistent gains over strong individual recallers and multi-channel fusion baselines, including up to 15.88% relative improvement in NDCG@50 on Food. Our implementation is available on https://github.com/zhengzaiyi/RoutePO.</abstract>
<identifier type="citekey">zheng-etal-2026-routepo</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.784/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>16593</start>
<end>16609</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T RoutePO: Personalized Routing Policy Optimization for LLM-based Multi-Channel Recall
%A Zheng, Zaiyi
%A Zhu, Yaochen
%A Ju, Mingxuan
%A Zhao, Tong
%A Shah, Neil
%A Li, Jundong
%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 zheng-etal-2026-routepo
%X LLM-based rerankers in recommender systems are inherently constrained by limited context windows: they can only reorder a small retrieved candidate set, so any relevant item missed by upstream retrieval cannot be recovered at ranking time. To mitigate this dependency, we propose RoutePO, which repositions the LLM from an item ranker to a recall strategy router for multi-channel recall. Conditioned on a user’s profile/context (interaction history and optional demographics), the LLM outputs personalized fusion weights over heterogeneous retrieval experts, enabling dynamic candidate aggregation without exposing long candidate lists from the entire item pool to the prompt. We train the router via a synergistic two-stage alignment process: Coarse Routing Alignment (SFT) warm-starts the policy with a quality-aware oracle prior over retrieval experts, and Fine-grained Metric Optimization (RL) directly optimizes recall-stage ranking metrics (e.g., NDCG) by sampling simplex-valued routing weights and rewarding fused-list quality. To support continuous exploration over the simplex under GRPO-style optimization, we introduce SofT-GRPO based on Gumbel–Softmax reparameterization with nucleus restriction and KL regularization. Finally, we instantiate fusion with both weighted score merging and a weight-aware interleaving strategy (Snake Fusion) that preserves within-channel rankings. Experiments on three real-world datasets demonstrate consistent gains over strong individual recallers and multi-channel fusion baselines, including up to 15.88% relative improvement in NDCG@50 on Food. Our implementation is available on https://github.com/zhengzaiyi/RoutePO.
%U https://aclanthology.org/2026.findings-emnlp.784/
%P 16593-16609
Markdown (Informal)
[RoutePO: Personalized Routing Policy Optimization for LLM-based Multi-Channel Recall](https://aclanthology.org/2026.findings-emnlp.784/) (Zheng et al., Findings 2026)
ACL