@inproceedings{park-etal-2025-maporl,
title = "{MAP}o{RL}: Multi-Agent Post-Co-Training for Collaborative Large Language Models with Reinforcement Learning",
author = "Park, Chanwoo and
Han, Seungju and
Guo, Xingzhi and
Ozdaglar, Asuman E. and
Zhang, Kaiqing and
Kim, Joo-Kyung",
editor = "Che, Wanxiang and
Nabende, Joyce and
Shutova, Ekaterina and
Pilehvar, Mohammad Taher",
booktitle = "Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2025",
address = "Vienna, Austria",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.acl-long.1459/",
doi = "10.18653/v1/2025.acl-long.1459",
pages = "30215--30248",
ISBN = "979-8-89176-251-0",
abstract = "Leveraging multi-agentic frameworks to enhance large language models (LLMs) has demonstrated significant potential recently, with most existing studies focusing on prompting and developing workflows with frozen LLMs. In this paper, we aim to further unleash the power of such multi-agentic frameworks for post-training LLMs for better collaboration. Specifically, we develop a new paradigm of Multi-Agent Post-co-training for collaborative LLMs with Reinforcement Learning (MAPoRL). In MAPoRL, multiple LLMs first generate their own responses and engage in discussions to collaboratively enhance the final response output; the final output is then scored by a verifier, where the scores serve as the reward and is maximized through multi-agent RL. Additionally, MAPoRL also reshapes the reward above with additional incentives to encourage corrective and persuasive outputs in the discussions. A key novelty from most existing LLM post-training paradigms is the advocacy of co-training multiple LLMs together, and the use of RL for better generalization. Accompanied by a few analytical insights, our experiments show that training single LLMs solely is insufficient for encouraging collaboration, while multi-agent co-training can significantly enhance the collaboration performance across multiple datasets, with generalization to unseen domains, compared to that of multiple LLMs before post-training."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="park-etal-2025-maporl">
<titleInfo>
<title>MAPoRL: Multi-Agent Post-Co-Training for Collaborative Large Language Models with Reinforcement Learning</title>
</titleInfo>
<name type="personal">
<namePart type="given">Chanwoo</namePart>
<namePart type="family">Park</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Seungju</namePart>
<namePart type="family">Han</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xingzhi</namePart>
<namePart type="family">Guo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Asuman</namePart>
<namePart type="given">E</namePart>
<namePart type="family">Ozdaglar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kaiqing</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Joo-Kyung</namePart>
<namePart type="family">Kim</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Wanxiang</namePart>
<namePart type="family">Che</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Joyce</namePart>
<namePart type="family">Nabende</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ekaterina</namePart>
<namePart type="family">Shutova</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mohammad</namePart>
<namePart type="given">Taher</namePart>
<namePart type="family">Pilehvar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Vienna, Austria</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-251-0</identifier>
</relatedItem>
<abstract>Leveraging multi-agentic frameworks to enhance large language models (LLMs) has demonstrated significant potential recently, with most existing studies focusing on prompting and developing workflows with frozen LLMs. In this paper, we aim to further unleash the power of such multi-agentic frameworks for post-training LLMs for better collaboration. Specifically, we develop a new paradigm of Multi-Agent Post-co-training for collaborative LLMs with Reinforcement Learning (MAPoRL). In MAPoRL, multiple LLMs first generate their own responses and engage in discussions to collaboratively enhance the final response output; the final output is then scored by a verifier, where the scores serve as the reward and is maximized through multi-agent RL. Additionally, MAPoRL also reshapes the reward above with additional incentives to encourage corrective and persuasive outputs in the discussions. A key novelty from most existing LLM post-training paradigms is the advocacy of co-training multiple LLMs together, and the use of RL for better generalization. Accompanied by a few analytical insights, our experiments show that training single LLMs solely is insufficient for encouraging collaboration, while multi-agent co-training can significantly enhance the collaboration performance across multiple datasets, with generalization to unseen domains, compared to that of multiple LLMs before post-training.</abstract>
<identifier type="citekey">park-etal-2025-maporl</identifier>
<identifier type="doi">10.18653/v1/2025.acl-long.1459</identifier>
<location>
<url>https://aclanthology.org/2025.acl-long.1459/</url>
</location>
<part>
<date>2025-07</date>
<extent unit="page">
<start>30215</start>
<end>30248</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T MAPoRL: Multi-Agent Post-Co-Training for Collaborative Large Language Models with Reinforcement Learning
%A Park, Chanwoo
%A Han, Seungju
%A Guo, Xingzhi
%A Ozdaglar, Asuman E.
%A Zhang, Kaiqing
%A Kim, Joo-Kyung
%Y Che, Wanxiang
%Y Nabende, Joyce
%Y Shutova, Ekaterina
%Y Pilehvar, Mohammad Taher
%S Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
%D 2025
%8 July
%I Association for Computational Linguistics
%C Vienna, Austria
%@ 979-8-89176-251-0
%F park-etal-2025-maporl
%X Leveraging multi-agentic frameworks to enhance large language models (LLMs) has demonstrated significant potential recently, with most existing studies focusing on prompting and developing workflows with frozen LLMs. In this paper, we aim to further unleash the power of such multi-agentic frameworks for post-training LLMs for better collaboration. Specifically, we develop a new paradigm of Multi-Agent Post-co-training for collaborative LLMs with Reinforcement Learning (MAPoRL). In MAPoRL, multiple LLMs first generate their own responses and engage in discussions to collaboratively enhance the final response output; the final output is then scored by a verifier, where the scores serve as the reward and is maximized through multi-agent RL. Additionally, MAPoRL also reshapes the reward above with additional incentives to encourage corrective and persuasive outputs in the discussions. A key novelty from most existing LLM post-training paradigms is the advocacy of co-training multiple LLMs together, and the use of RL for better generalization. Accompanied by a few analytical insights, our experiments show that training single LLMs solely is insufficient for encouraging collaboration, while multi-agent co-training can significantly enhance the collaboration performance across multiple datasets, with generalization to unseen domains, compared to that of multiple LLMs before post-training.
%R 10.18653/v1/2025.acl-long.1459
%U https://aclanthology.org/2025.acl-long.1459/
%U https://doi.org/10.18653/v1/2025.acl-long.1459
%P 30215-30248
Markdown (Informal)
[MAPoRL: Multi-Agent Post-Co-Training for Collaborative Large Language Models with Reinforcement Learning](https://aclanthology.org/2025.acl-long.1459/) (Park et al., ACL 2025)
ACL