CoReaAgents: A Collaboration and Reasoning Framework Based on LLM-Powered Agents for Complex Reasoning Tasks

As LLMs demonstrate remarkable reasoning capabilities, LLM-powered agents are seen as key to achieving AGI (Artificial General Intelligence) and are widely applied in various complex real-world scenarios. Nevertheless, existing studies still suffer from missing steps, deviated task execution and inc...

Full description

Saved in:
Bibliographic Details
Main Authors: Zhonghe Han, Jiaxin Wang, Xiaolu Yan, Zhiying Jiang, Yuanben Zhang, Siye Liu, Qihang Gong, Chenwei Song
Format: Article
Language:English
Published: MDPI AG 2025-05-01
Series:Applied Sciences
Subjects:
Online Access:https://www.mdpi.com/2076-3417/15/10/5663
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849711101129785344
author Zhonghe Han
Jiaxin Wang
Xiaolu Yan
Zhiying Jiang
Yuanben Zhang
Siye Liu
Qihang Gong
Chenwei Song
author_facet Zhonghe Han
Jiaxin Wang
Xiaolu Yan
Zhiying Jiang
Yuanben Zhang
Siye Liu
Qihang Gong
Chenwei Song
author_sort Zhonghe Han
collection DOAJ
description As LLMs demonstrate remarkable reasoning capabilities, LLM-powered agents are seen as key to achieving AGI (Artificial General Intelligence) and are widely applied in various complex real-world scenarios. Nevertheless, existing studies still suffer from missing steps, deviated task execution and incorrect tool selection. This paper proposes CoReaAgents, a collaboration and reasoning framework based on LLM-powered agents, comprising the Plan Agent (as a precise task planner), the Tool Agent (as a proficient tool user) and the Reflect Agent (as an objective task evaluator). These agents simulate the social division of labor and synergistic cooperation to enable each agent to perform different specialized capabilities in order to solve complex tasks together. Through the above mechanism, the CoReaAgents framework has the skills of prospective thinking and flexible execution. To verify the capability of the CoReaAgents framework, this paper conducts extensive experiments on different complex tasks such as tool learning, math reasoning and multi-hop QA. The results show that the CoReaAgents framework outperforms various comparative methods in both quantity and quality.
format Article
id doaj-art-11e6d863e9e24b82a9e0f269ee9cc048
institution DOAJ
issn 2076-3417
language English
publishDate 2025-05-01
publisher MDPI AG
record_format Article
series Applied Sciences
spelling doaj-art-11e6d863e9e24b82a9e0f269ee9cc0482025-08-20T03:14:42ZengMDPI AGApplied Sciences2076-34172025-05-011510566310.3390/app15105663CoReaAgents: A Collaboration and Reasoning Framework Based on LLM-Powered Agents for Complex Reasoning TasksZhonghe Han0Jiaxin Wang1Xiaolu Yan2Zhiying Jiang3Yuanben Zhang4Siye Liu5Qihang Gong6Chenwei Song7Aerospace Information Research Institute, Chinese Academy of Sciences, Beijing 100190, ChinaCollege of Information Science and Technology, Beijing University of Chemical Technology, Beijing 100029, ChinaInstitute of Tracking and Telecommunications Technology, Beijing 100094, ChinaCollege of Information Science and Technology, Beijing University of Chemical Technology, Beijing 100029, ChinaAerospace Information Research Institute, Chinese Academy of Sciences, Beijing 100190, ChinaAerospace Information Research Institute, Chinese Academy of Sciences, Beijing 100190, ChinaAerospace Information Research Institute, Chinese Academy of Sciences, Beijing 100190, ChinaCollege of Information Science and Technology, Beijing University of Chemical Technology, Beijing 100029, ChinaAs LLMs demonstrate remarkable reasoning capabilities, LLM-powered agents are seen as key to achieving AGI (Artificial General Intelligence) and are widely applied in various complex real-world scenarios. Nevertheless, existing studies still suffer from missing steps, deviated task execution and incorrect tool selection. This paper proposes CoReaAgents, a collaboration and reasoning framework based on LLM-powered agents, comprising the Plan Agent (as a precise task planner), the Tool Agent (as a proficient tool user) and the Reflect Agent (as an objective task evaluator). These agents simulate the social division of labor and synergistic cooperation to enable each agent to perform different specialized capabilities in order to solve complex tasks together. Through the above mechanism, the CoReaAgents framework has the skills of prospective thinking and flexible execution. To verify the capability of the CoReaAgents framework, this paper conducts extensive experiments on different complex tasks such as tool learning, math reasoning and multi-hop QA. The results show that the CoReaAgents framework outperforms various comparative methods in both quantity and quality.https://www.mdpi.com/2076-3417/15/10/5663LLMLLM-powered agentartificial intelligencemulti-agent systemsmachine learningsmart tools and applications
spellingShingle Zhonghe Han
Jiaxin Wang
Xiaolu Yan
Zhiying Jiang
Yuanben Zhang
Siye Liu
Qihang Gong
Chenwei Song
CoReaAgents: A Collaboration and Reasoning Framework Based on LLM-Powered Agents for Complex Reasoning Tasks
Applied Sciences
LLM
LLM-powered agent
artificial intelligence
multi-agent systems
machine learning
smart tools and applications
title CoReaAgents: A Collaboration and Reasoning Framework Based on LLM-Powered Agents for Complex Reasoning Tasks
title_full CoReaAgents: A Collaboration and Reasoning Framework Based on LLM-Powered Agents for Complex Reasoning Tasks
title_fullStr CoReaAgents: A Collaboration and Reasoning Framework Based on LLM-Powered Agents for Complex Reasoning Tasks
title_full_unstemmed CoReaAgents: A Collaboration and Reasoning Framework Based on LLM-Powered Agents for Complex Reasoning Tasks
title_short CoReaAgents: A Collaboration and Reasoning Framework Based on LLM-Powered Agents for Complex Reasoning Tasks
title_sort coreaagents a collaboration and reasoning framework based on llm powered agents for complex reasoning tasks
topic LLM
LLM-powered agent
artificial intelligence
multi-agent systems
machine learning
smart tools and applications
url https://www.mdpi.com/2076-3417/15/10/5663
work_keys_str_mv AT zhonghehan coreaagentsacollaborationandreasoningframeworkbasedonllmpoweredagentsforcomplexreasoningtasks
AT jiaxinwang coreaagentsacollaborationandreasoningframeworkbasedonllmpoweredagentsforcomplexreasoningtasks
AT xiaoluyan coreaagentsacollaborationandreasoningframeworkbasedonllmpoweredagentsforcomplexreasoningtasks
AT zhiyingjiang coreaagentsacollaborationandreasoningframeworkbasedonllmpoweredagentsforcomplexreasoningtasks
AT yuanbenzhang coreaagentsacollaborationandreasoningframeworkbasedonllmpoweredagentsforcomplexreasoningtasks
AT siyeliu coreaagentsacollaborationandreasoningframeworkbasedonllmpoweredagentsforcomplexreasoningtasks
AT qihanggong coreaagentsacollaborationandreasoningframeworkbasedonllmpoweredagentsforcomplexreasoningtasks
AT chenweisong coreaagentsacollaborationandreasoningframeworkbasedonllmpoweredagentsforcomplexreasoningtasks