A Multiagent Reasoning Framework for Classical Chinese Question Answering With Large Language Models.
Saved in:
| Title: | A Multiagent Reasoning Framework for Classical Chinese Question Answering With Large Language Models. |
|---|---|
| Authors: | Liu, Qing1 (AUTHOR) liuqing@tfswufe.edu.cn, Nong, Bin1 (AUTHOR), Jia, Xuemei1 (AUTHOR), Zhang, Yajie2 (AUTHOR), Chen, Hongmeng3 (AUTHOR), Deng, Yulin1 (AUTHOR), Murray, Richard (AUTHOR) rmurray@wiley.com |
| Source: | International Journal of Intelligent Systems. 4/28/2026, Vol. 2026, p1-13. 13p. |
| Subjects: | Language models, Intelligent tutoring systems, Educational technology, Natural language processing, Chinese literature, Question & answer websites |
| Abstract: | Background: Understanding Classical Chinese remains a major challenge in Chinese education, especially in the National College Entrance Examination (NCEE). Although large language models (LLMs) exhibit strong reasoning capabilities, their performance on exam‐style Classical Chinese questions still suffers from instability and limited accuracy. Methods: We propose a multiagent reasoning framework based on LLMs for Classical Chinese question answering. For each question type, standardized reasoning procedures are defined, and specialized agents are trained for subtasks including word interpretation, grammatical analysis, translation, and semantic summarization. A two‐round reasoning mechanism, consisting of an initial response followed by refinement using standard answers, is introduced to enhance consistency and robustness. Results: Experiments on Gaokao‐style Classical Chinese questions demonstrate that the proposed framework achieves higher accuracy and greater reasoning stability than single‐agent systems and general‐purpose LLMs. In objective tasks, it outperforms strong Chinese‐oriented models such as Qwen‐Max and Baichuan‐4 by up to 6.8%. Conclusions: The proposed multiagent framework improves both the interpretability and reliability of LLM‐based Classical Chinese understanding. It shows strong potential for applications in intelligent tutoring systems, curriculum support, and cognitive modeling of human‐like reasoning in educational contexts. [ABSTRACT FROM AUTHOR] |
| Copyright of International Journal of Intelligent Systems is the property of Wiley-Blackwell and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.) | |
| Database: | Engineering Source |
|
Full text is not displayed to guests.
Login for full access.
|
|
| Abstract: | Background: Understanding Classical Chinese remains a major challenge in Chinese education, especially in the National College Entrance Examination (NCEE). Although large language models (LLMs) exhibit strong reasoning capabilities, their performance on exam‐style Classical Chinese questions still suffers from instability and limited accuracy. Methods: We propose a multiagent reasoning framework based on LLMs for Classical Chinese question answering. For each question type, standardized reasoning procedures are defined, and specialized agents are trained for subtasks including word interpretation, grammatical analysis, translation, and semantic summarization. A two‐round reasoning mechanism, consisting of an initial response followed by refinement using standard answers, is introduced to enhance consistency and robustness. Results: Experiments on Gaokao‐style Classical Chinese questions demonstrate that the proposed framework achieves higher accuracy and greater reasoning stability than single‐agent systems and general‐purpose LLMs. In objective tasks, it outperforms strong Chinese‐oriented models such as Qwen‐Max and Baichuan‐4 by up to 6.8%. Conclusions: The proposed multiagent framework improves both the interpretability and reliability of LLM‐based Classical Chinese understanding. It shows strong potential for applications in intelligent tutoring systems, curriculum support, and cognitive modeling of human‐like reasoning in educational contexts. [ABSTRACT FROM AUTHOR] |
|---|---|
| ISSN: | 08848173 |
| DOI: | 10.1155/int/8987020 |
