Research paper

CoEvoLoop: Closed-Loop RL Training of Customer-Service Agents by Co-Evolving Environment, Reward, and Policy

Published by Association for Computational LinguisticsOct 2026PDF

Explore a training loop that adapts simulated environments and rewards alongside a customer service agent.

Customer support Model evaluation

CoEvoLoop: Closed-Loop RL Training of Customer-Service Agents by Co-Evolving Environment, Reward, and Policy cover

Resource access

Get the full resource

Enter your business details to read the full resource.

First nameLast nameWork emailCompanyJob titleCountry or region

Hui Cai, Jianda Chen, Yuhong Zhou, Wei Ren, Choy Zhen Wen Marcus, Tan Li Hui Faith, Lyuxin Xue, Hao Chen, Jiawei Luo, Xuanwei Hu, Xin Zhang. CoEvoLoop: Closed-Loop RL Training of Customer-Service Agents by Co-Evolving Environment, Reward, and Policy. Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing: System Demonstrations. © 2026 Association for Computational Linguistics. CC BY 4.0. 10.18653/2026.emnlp-demos.24. Original PDF, unmodified.

CC BY 4.0

CoEvoLoop: Closed-Loop RL Training of Customer-Service Agents by Co-Evolving Environment, Reward, and Policy

Related research.

Learning Selective LLM Autonomy from Copilot Feedback in Enterprise Customer Support Workflows

ECom-Bench: Can LLM Agent Resolve Real-World E-commerce Customer Support Issues?

Retrieval Enhancements for RAG: Insights from a Deployed Customer Support Chatbot