@inproceedings{zhang-zhao-2025-collaborative,
title = "A Collaborative Reasoning Framework Powered by Reinforcement Learning and Large Language Models for Complex Questions Answering over Knowledge Graph",
author = "Zhang, Zhiqiang and
Zhao, Wen",
editor = "Rambow, Owen and
Wanner, Leo and
Apidianaki, Marianna and
Al-Khalifa, Hend and
Eugenio, Barbara Di and
Schockaert, Steven",
booktitle = "Proceedings of the 31st International Conference on Computational Linguistics",
month = jan,
year = "2025",
address = "Abu Dhabi, UAE",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.coling-main.712/",
pages = "10672--10684",
abstract = "Knowledge Graph Question Answering (KGQA) aims to automatically answer natural language questions by reasoning across multiple triples in knowledge graphs (KGs). Reinforcement learning (RL)-based methods are introduced to enhance model interpretability. Nevertheless, when addressing complex questions requiring long-term reasoning, the RL agent is usually misled by aimless exploration, as it lacks common learning practices with prior knowledge. Recently, large language models (LLMs) have been proven to encode vast amounts of knowledge about the world and possess remarkable reasoning capabilities. However, they often encounter challenges with hallucination issues, failing to address complex questions that demand deep and deliberate reasoning. In this paper, we propose a collaborative reasoning framework (CRF) powered by RL and LLMs to answer complex questions based on the knowledge graph. Our approach leverages the common sense priors contained in LLMs while utilizing RL to provide learning from the environment, resulting in a hierarchical agent that uses LLMs to solve the complex KGQA task. By combining LLMs and the RL policy, the high-level agent accurately identifies constraints encountered during reasoning, while the low-level agent conducts efficient path reasoning by selecting the most promising relations in KG. Extensive experiments conducted on four benchmark datasets clearly demonstrate the effectiveness of the proposed model, which surpasses state-of-the-art approaches."
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="zhang-zhao-2025-collaborative">
<titleInfo>
<title>A Collaborative Reasoning Framework Powered by Reinforcement Learning and Large Language Models for Complex Questions Answering over Knowledge Graph</title>
</titleInfo>
<name type="personal">
<namePart type="given">Zhiqiang</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Wen</namePart>
<namePart type="family">Zhao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-01</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 31st International Conference on Computational Linguistics</title>
</titleInfo>
<name type="personal">
<namePart type="given">Owen</namePart>
<namePart type="family">Rambow</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Leo</namePart>
<namePart type="family">Wanner</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Marianna</namePart>
<namePart type="family">Apidianaki</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hend</namePart>
<namePart type="family">Al-Khalifa</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Barbara</namePart>
<namePart type="given">Di</namePart>
<namePart type="family">Eugenio</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Steven</namePart>
<namePart type="family">Schockaert</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Abu Dhabi, UAE</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>Knowledge Graph Question Answering (KGQA) aims to automatically answer natural language questions by reasoning across multiple triples in knowledge graphs (KGs). Reinforcement learning (RL)-based methods are introduced to enhance model interpretability. Nevertheless, when addressing complex questions requiring long-term reasoning, the RL agent is usually misled by aimless exploration, as it lacks common learning practices with prior knowledge. Recently, large language models (LLMs) have been proven to encode vast amounts of knowledge about the world and possess remarkable reasoning capabilities. However, they often encounter challenges with hallucination issues, failing to address complex questions that demand deep and deliberate reasoning. In this paper, we propose a collaborative reasoning framework (CRF) powered by RL and LLMs to answer complex questions based on the knowledge graph. Our approach leverages the common sense priors contained in LLMs while utilizing RL to provide learning from the environment, resulting in a hierarchical agent that uses LLMs to solve the complex KGQA task. By combining LLMs and the RL policy, the high-level agent accurately identifies constraints encountered during reasoning, while the low-level agent conducts efficient path reasoning by selecting the most promising relations in KG. Extensive experiments conducted on four benchmark datasets clearly demonstrate the effectiveness of the proposed model, which surpasses state-of-the-art approaches.</abstract>
<identifier type="citekey">zhang-zhao-2025-collaborative</identifier>
<location>
<url>https://aclanthology.org/2025.coling-main.712/</url>
</location>
<part>
<date>2025-01</date>
<extent unit="page">
<start>10672</start>
<end>10684</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T A Collaborative Reasoning Framework Powered by Reinforcement Learning and Large Language Models for Complex Questions Answering over Knowledge Graph
%A Zhang, Zhiqiang
%A Zhao, Wen
%Y Rambow, Owen
%Y Wanner, Leo
%Y Apidianaki, Marianna
%Y Al-Khalifa, Hend
%Y Eugenio, Barbara Di
%Y Schockaert, Steven
%S Proceedings of the 31st International Conference on Computational Linguistics
%D 2025
%8 January
%I Association for Computational Linguistics
%C Abu Dhabi, UAE
%F zhang-zhao-2025-collaborative
%X Knowledge Graph Question Answering (KGQA) aims to automatically answer natural language questions by reasoning across multiple triples in knowledge graphs (KGs). Reinforcement learning (RL)-based methods are introduced to enhance model interpretability. Nevertheless, when addressing complex questions requiring long-term reasoning, the RL agent is usually misled by aimless exploration, as it lacks common learning practices with prior knowledge. Recently, large language models (LLMs) have been proven to encode vast amounts of knowledge about the world and possess remarkable reasoning capabilities. However, they often encounter challenges with hallucination issues, failing to address complex questions that demand deep and deliberate reasoning. In this paper, we propose a collaborative reasoning framework (CRF) powered by RL and LLMs to answer complex questions based on the knowledge graph. Our approach leverages the common sense priors contained in LLMs while utilizing RL to provide learning from the environment, resulting in a hierarchical agent that uses LLMs to solve the complex KGQA task. By combining LLMs and the RL policy, the high-level agent accurately identifies constraints encountered during reasoning, while the low-level agent conducts efficient path reasoning by selecting the most promising relations in KG. Extensive experiments conducted on four benchmark datasets clearly demonstrate the effectiveness of the proposed model, which surpasses state-of-the-art approaches.
%U https://aclanthology.org/2025.coling-main.712/
%P 10672-10684
Markdown (Informal)
[A Collaborative Reasoning Framework Powered by Reinforcement Learning and Large Language Models for Complex Questions Answering over Knowledge Graph](https://aclanthology.org/2025.coling-main.712/) (Zhang & Zhao, COLING 2025)
ACL