@inproceedings{cai-etal-2026-coevoloop,
title = "{C}o{E}vo{L}oop: Closed-Loop {RL} Training of Customer-Service Agents by Co-Evolving Environment, Reward, and Policy",
author = "Cai, Hui and
Chen, Jianda and
Zhou, Yuhong and
Ren, Wei and
Marcus, Choy Zhen Wen and
Faith, Tan Li Hui and
Xue, Lyuxin and
Chen, Hao and
Luo, Jiawei and
Hu, Xuanwei and
Zhang, Xin",
editor = "Agrawal, Sweta and
Abad, Alberto",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing: System Demonstrations",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-demos.24/",
pages = "242--252",
ISBN = "979-8-89176-461-3",
abstract = "Developing enterprise customer-service assistants with Reinforcement Learning (RL) critically depends on high-fidelity user simulators and well-aligned reward functions, since real-customer interactions are impractical for large-scale training and rule-based rewards fail to capture complex business requirements. We present CoEvoLoop, a framework that co-evolves the assistant policy together with these RL components through an expert-calibrated, multi-round loop. Beyond an LLM-based simulator for rollout generation and a multi-dimensional grader for rubric-based evaluation of service quality and compliance, CoEvoLoop integrates an expert annotation interface where sampled trajectories and grader judgments are reviewed and corrected. Each round, expert feedback updates the simulator and grader, after which the assistant is optimized with trajectory-level RL against the updated evaluation signal; monitoring interfaces support rollout inspection, dimension-wise failure attribution, and cross-round tracking. In an industrial evaluation, the system improves multi-turn service quality and reduces compliance violations over training rounds; component ablations and blind expert evaluation confirm that co-evolving the simulator and grader is load-bearing for these gains. We release the system demo code, along with a demo video."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="cai-etal-2026-coevoloop">
<titleInfo>
<title>CoEvoLoop: Closed-Loop RL Training of Customer-Service Agents by Co-Evolving Environment, Reward, and Policy</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hui</namePart>
<namePart type="family">Cai</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jianda</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yuhong</namePart>
<namePart type="family">Zhou</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Wei</namePart>
<namePart type="family">Ren</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Choy</namePart>
<namePart type="given">Zhen</namePart>
<namePart type="given">Wen</namePart>
<namePart type="family">Marcus</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tan</namePart>
<namePart type="given">Li</namePart>
<namePart type="given">Hui</namePart>
<namePart type="family">Faith</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Lyuxin</namePart>
<namePart type="family">Xue</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hao</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiawei</namePart>
<namePart type="family">Luo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xuanwei</namePart>
<namePart type="family">Hu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xin</namePart>
<namePart type="family">Zhang</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: System Demonstrations</title>
</titleInfo>
<name type="personal">
<namePart type="given">Sweta</namePart>
<namePart type="family">Agrawal</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Alberto</namePart>
<namePart type="family">Abad</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">979-8-89176-461-3</identifier>
</relatedItem>
<abstract>Developing enterprise customer-service assistants with Reinforcement Learning (RL) critically depends on high-fidelity user simulators and well-aligned reward functions, since real-customer interactions are impractical for large-scale training and rule-based rewards fail to capture complex business requirements. We present CoEvoLoop, a framework that co-evolves the assistant policy together with these RL components through an expert-calibrated, multi-round loop. Beyond an LLM-based simulator for rollout generation and a multi-dimensional grader for rubric-based evaluation of service quality and compliance, CoEvoLoop integrates an expert annotation interface where sampled trajectories and grader judgments are reviewed and corrected. Each round, expert feedback updates the simulator and grader, after which the assistant is optimized with trajectory-level RL against the updated evaluation signal; monitoring interfaces support rollout inspection, dimension-wise failure attribution, and cross-round tracking. In an industrial evaluation, the system improves multi-turn service quality and reduces compliance violations over training rounds; component ablations and blind expert evaluation confirm that co-evolving the simulator and grader is load-bearing for these gains. We release the system demo code, along with a demo video.</abstract>
<identifier type="citekey">cai-etal-2026-coevoloop</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-demos.24/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>242</start>
<end>252</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T CoEvoLoop: Closed-Loop RL Training of Customer-Service Agents by Co-Evolving Environment, Reward, and Policy
%A Cai, Hui
%A Chen, Jianda
%A Zhou, Yuhong
%A Ren, Wei
%A Marcus, Choy Zhen Wen
%A Faith, Tan Li Hui
%A Xue, Lyuxin
%A Chen, Hao
%A Luo, Jiawei
%A Hu, Xuanwei
%A Zhang, Xin
%Y Agrawal, Sweta
%Y Abad, Alberto
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing: System Demonstrations
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-461-3
%F cai-etal-2026-coevoloop
%X Developing enterprise customer-service assistants with Reinforcement Learning (RL) critically depends on high-fidelity user simulators and well-aligned reward functions, since real-customer interactions are impractical for large-scale training and rule-based rewards fail to capture complex business requirements. We present CoEvoLoop, a framework that co-evolves the assistant policy together with these RL components through an expert-calibrated, multi-round loop. Beyond an LLM-based simulator for rollout generation and a multi-dimensional grader for rubric-based evaluation of service quality and compliance, CoEvoLoop integrates an expert annotation interface where sampled trajectories and grader judgments are reviewed and corrected. Each round, expert feedback updates the simulator and grader, after which the assistant is optimized with trajectory-level RL against the updated evaluation signal; monitoring interfaces support rollout inspection, dimension-wise failure attribution, and cross-round tracking. In an industrial evaluation, the system improves multi-turn service quality and reduces compliance violations over training rounds; component ablations and blind expert evaluation confirm that co-evolving the simulator and grader is load-bearing for these gains. We release the system demo code, along with a demo video.
%U https://aclanthology.org/2026.emnlp-demos.24/
%P 242-252
Markdown (Informal)
[CoEvoLoop: Closed-Loop RL Training of Customer-Service Agents by Co-Evolving Environment, Reward, and Policy](https://aclanthology.org/2026.emnlp-demos.24/) (Cai et al., EMNLP 2026)
ACL
- Hui Cai, Jianda Chen, Yuhong Zhou, Wei Ren, Choy Zhen Wen Marcus, Tan Li Hui Faith, Lyuxin Xue, Hao Chen, Jiawei Luo, Xuanwei Hu, and Xin Zhang. 2026. CoEvoLoop: Closed-Loop RL Training of Customer-Service Agents by Co-Evolving Environment, Reward, and Policy. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing: System Demonstrations, pages 242–252, Budapest, Hungary. Association for Computational Linguistics.