Article(id=1329420562385973463, tenantId=1146029695717560320, journalId=1324759914146529478, issueId=1329420435915125316, articleNumber=null, orderNo=null, doi=10.7544/issn1000-1239.202660421, pmid=null, cstr=null, oa=null, hot=null, price=null, onlineType=0, articleFormat=0, articleType=null, articleTypeStr=null, receivedDate=1780675200000, receivedDateStr=2026-06-06, revisedDate=1784044800000, revisedDateStr=2026-07-15, acceptedDate=null, acceptedDateStr=null, onlineDate=1794888161511, onlineDateStr=2026-11-17, pubDate=1788192000000, pubDateStr=2026-09-01, doiRegisterDate=null, doiRegisterDateStr=null, onlineIssueDate=1794888161511, onlineIssueDateStr=2026-11-17, onlineJustAcceptDate=null, onlineJustAcceptDateStr=null, onlineFirstDate=null, onlineFirstDateStr=null, sourceXml=null, magXml=null, createTime=1794888161511, creator=13701087609, updateTime=1794888161511, updator=13701087609, issue=Issue{id=1329420435915125316, tenantId=1146029695717560320, journalId=1324759914146529478, year='2026', volume='63', issue='9', pageStart='2183', pageEnd='2477', issueExtLink='null', onlineDate='null', pubDate='1788192000000', pubDateStr='2026-09-01', beforeIssueId=null, nextIssueId=null, price=null, status=1, issueComplete=1, articleOrder=1, issueType=-1, specialIssue=null, createTime=1794888131359, creator='13701087609', updateTime=1794888131359, updator='13701087609', preIssue=null, nextIssue=null, articleTotal=null, ext=null, issueFiles=null, downloadFileDto=null}, startPage=2319, endPage=2337, ext={EN=ArticleExt(id=1329420562801209560, articleId=1329420562385973463, tenantId=1146029695717560320, journalId=1324759914146529478, language=EN, title=CAKT-Agent: A Context-Adaptive Knowledge Tracing Agent Based on MoE, columnId=1329420436590408262, journalTitle=Journal of Computer Research and Development, columnName=Special Issue on Frontier Technologies and Applications of Knowledge-Enhanced Artificial Intelligence, runingTitle=null, highlight=null, articleAbstract=

Knowledge tracing, as the core task of intelligent tutoring systems, aims to predict students’ future learning performance by analyzing their historical behavioral data, and provides a fundamental basis for personalized learning path planning. Current mainstream deep learning methods are constrained by knowledge boundaries and rely on single-feature modeling, often exhibiting issues such as overly simplified learning process modeling, inadequate contextual adaptability, and black-box prediction mechanisms that yield results lacking interpretability and difficult to translate into actionable teaching references. To address these challenges, we propose CAKT-Agent, a context-adaptive knowledge tracing agent based on mixture-of-experts model, and construct a three-tier collaborative architecture consisting of a unified memory module, a learning agent, and a companion agent: the unified memory module realizes global storage and sharing of learning data, educational priors and intermediate model states; The learning agent defines a five-dimensional context based on educational measurement theory and cognitive science, designs a hybrid expert architecture with dimension-aware capabilities, and achieves context-adaptive integration through a two-level gating mechanism; the companion agent generates learning diagnostic reports with educational comprehensibility based on expert activation weights and educational theory rules. Experimental results on real-world datasets show that the proposed model outperforms mainstream knowledge tracing models in the prediction accuracy of students' learning performance, with excellent context-adaptive capability and result comprehensibility.

, authors=Jiaqi Long1, Suojuan Zhang1, *, Wenning Hao1, Zejun Chen1, Xinjie Sun2, 3, Yifan Liu1, Enhong Chen2, authorsList=Jiaqi Long, Suojuan Zhang, Wenning Hao, Zejun Chen, Xinjie Sun, Yifan Liu, Enhong Chen, authorCompany=null, correspAuthors=Suojuan Zhang, authorNote=

born in 2002. Master candidate. Her main research interests include intelligent education and knowledge tracing

born in 1983. PhD, professor. Her main research interests include educational data mining, knowledge tracing, and personalized exercise recommendation

born in 1971. PhD, professor. His main research interests include large-scale high-dimensional data compression and evaluation of operational effectiveness

born in 1998. Master candidate. Her main research interests include intelligent education and exercise recommendation

born in 1987. PhD, professor. His main research interests include intelligent education and adaptive learning

born in 2000. Master candidate. His main research interests include intelligent education and data mining

born in 1968. PhD, professor. His main research interests include machine learning, data mining, education big data analysis, and personalized recommendation system

, correspAuthorsNote=null, copyrightStatement=Copyright © 2026 Journal of Computer Research and Development. All rights reserved., copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=null, magXml=null, pdfUrl=null, pdf=null, pdfFileSize=null, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=null, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=null, mapNumber=null, fund=null), CN=ArticleExt(id=1329420565506535671, articleId=1329420562385973463, tenantId=1146029695717560320, journalId=1324759914146529478, language=CN, title=CAKT-Agent:基于MoE的情境自适应知识追踪智能体, columnId=1329420436737208904, journalTitle=计算机研究与发展, columnName=“知识增强的人工智能前沿技术及应用”专题, runingTitle=null, highlight=null, articleAbstract=

知识追踪作为智能辅导系统的核心任务,旨在通过分析学生历史行为数据预测其未来表现,为个性化学习路径规划提供依据。当前主流深度学习方法受限于知识边界,依赖单一特征拟合,普遍存在学习过程建模过度简化、情境适配能力不足、预测过程黑箱化等问题,输出结果缺乏可解释性,难以转化为可落地的教学参考。为此,提出了基于混合专家(mixture of experts,MoE)模型的情境自适应知识追踪智能体CAKT-Agent,构建统一记忆模块、学习智能体和伴学智能体三级协同架构:统一记忆模块实现学习数据、教育先验知识和中间状态的全局存储与共享;学习智能体基于教育测量学和认知科学定义5维情境,设计维度感知的混合专家架构,通过两级门控机制实现情境自适应融合;伴学智能体依托专家激活权重与教育理论规则,生成具备教育可理解性的学习诊断报告。真实数据集实验结果表明,所提模型在学生表现预测精度上优于主流模型,且具备良好的情境自适应能力与结果可理解性。

, authors=龙佳琪1, 张所娟1, *, 郝文宁1, 陈泽君1, 孙新杰2, 3, 刘逸凡1, 陈恩红2, authorsList=龙佳琪, 张所娟, 郝文宁, 陈泽君, 孙新杰, 刘逸凡, 陈恩红, authorCompany=null, correspAuthors=张所娟, authorNote=

作者贡献声明

龙佳琪负责模型设计并撰写论文;张所娟制定论文整体研究框架;郝文宁提出关键改进意见;陈泽君、孙新杰参与了部分实验方法的分析与讨论;刘逸凡协助完成了数据收集与初步整理工作;陈恩红统筹研究进度与团队协作,确定论文选题方向。

2002年生。硕士研究生。主要研究方向为智能教育、知识追踪

1983年生。博士,教授。主要研究方向为教育数据挖掘、知识追踪、个性化学习推荐

1971年生。博士,教授。主要研究方向为大规模高维数据压缩、作战效能评估

1998年生。硕士研究生。主要研究方向为智能教育、学习推荐

1987年生。博士,教授。主要研究方向为智能教育、自适应学习

2000年生。硕士研究生。主要研究方向为智能教育、数据挖掘

1968年生。博士,教授。主要研究方向为机器学习、数据挖掘、教育大数据分析、个性化推荐系统

, correspAuthorsNote=
张所娟()
, copyrightStatement=版权所有 © 2026 计算机研究与发展编辑部, copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=R5GStrzpMCxmxv2l/VdF8w==, magXml=9u//jKmUcRGKla1CZWFPfA==, pdfUrl=null, pdf=kahZKEqvmJViEugqLfapkA==, pdfFileSize=2708339, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=S0UuvqUXv2p7HXu+3PuXsQ==, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=szS15b3A3hd8FxF92lrsRQ==, mapNumber=null, fund=null)}, authors=[Author(id=1329420566156652802, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=jiaqilong@aeu.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329420566240538884, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, authorId=1329420566156652802, language=EN, stringName=Jiaqi Long, firstName=Jiaqi, middleName=null, lastName=Long, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1School of Command and Control Engineering, Army Engineering University of PLA, Nanjing 210007, bio={"content":"

born in 2002. Master candidate. Her main research interests include intelligent education and knowledge tracing

"}, bioImg=null, bioContent=

born in 2002. Master candidate. Her main research interests include intelligent education and knowledge tracing

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329420566316036357, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, authorId=1329420566156652802, language=CN, stringName=龙佳琪, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1中国人民解放军陆军工程大学指挥控制工程学院 南京 210007, bio={"img":"tYa6D1n98DG+b5+W73u8aA==","content":"

2002年生。硕士研究生。主要研究方向为智能教育、知识追踪

"}, bioImg=tYa6D1n98DG+b5+W73u8aA==, bioContent=

2002年生。硕士研究生。主要研究方向为智能教育、知识追踪

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329420565707862264, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, xref=1, ext=[AuthorCompanyExt(id=1329420565728833785, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, companyId=1329420565707862264, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1School of Command and Control Engineering, Army Engineering University of PLA, Nanjing 210007), AuthorCompanyExt(id=1329420565737222394, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, companyId=1329420565707862264, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1中国人民解放军陆军工程大学指挥控制工程学院 南京 210007)])]), Author(id=1329420566387339527, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, orderNo=1, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=suojuanzhang@aeu.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=1, authorType=1, ext={EN=AuthorExt(id=1329420566479614217, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, authorId=1329420566387339527, language=EN, stringName=Suojuan Zhang, firstName=Suojuan, middleName=null, lastName=Zhang, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, *, address=1School of Command and Control Engineering, Army Engineering University of PLA, Nanjing 210007, bio={"content":"

born in 1983. PhD, professor. Her main research interests include educational data mining, knowledge tracing, and personalized exercise recommendation

"}, bioImg=null, bioContent=

born in 1983. PhD, professor. Her main research interests include educational data mining, knowledge tracing, and personalized exercise recommendation

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329420566571888906, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, authorId=1329420566387339527, language=CN, stringName=张所娟, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, *, address=1中国人民解放军陆军工程大学指挥控制工程学院 南京 210007, bio={"img":"kp7oy3Hv4nf/jLmTywDAyg==","content":"

1983年生。博士,教授。主要研究方向为教育数据挖掘、知识追踪、个性化学习推荐

"}, bioImg=kp7oy3Hv4nf/jLmTywDAyg==, bioContent=

1983年生。博士,教授。主要研究方向为教育数据挖掘、知识追踪、个性化学习推荐

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329420565707862264, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, xref=1, ext=[AuthorCompanyExt(id=1329420565728833785, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, companyId=1329420565707862264, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1School of Command and Control Engineering, Army Engineering University of PLA, Nanjing 210007), AuthorCompanyExt(id=1329420565737222394, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, companyId=1329420565707862264, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1中国人民解放军陆军工程大学指挥控制工程学院 南京 210007)])]), Author(id=1329420566630609164, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, orderNo=2, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329420566701912334, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, authorId=1329420566630609164, language=EN, stringName=Wenning Hao, firstName=Wenning, middleName=null, lastName=Hao, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1School of Command and Control Engineering, Army Engineering University of PLA, Nanjing 210007, bio={"content":"

born in 1971. PhD, professor. His main research interests include large-scale high-dimensional data compression and evaluation of operational effectiveness

"}, bioImg=null, bioContent=

born in 1971. PhD, professor. His main research interests include large-scale high-dimensional data compression and evaluation of operational effectiveness

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329420566769021199, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, authorId=1329420566630609164, language=CN, stringName=郝文宁, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1中国人民解放军陆军工程大学指挥控制工程学院 南京 210007, bio={"img":"5ATAy/bWzMHW9yGN3NEsVA==","content":"

1971年生。博士,教授。主要研究方向为大规模高维数据压缩、作战效能评估

"}, bioImg=5ATAy/bWzMHW9yGN3NEsVA==, bioContent=

1971年生。博士,教授。主要研究方向为大规模高维数据压缩、作战效能评估

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329420565707862264, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, xref=1, ext=[AuthorCompanyExt(id=1329420565728833785, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, companyId=1329420565707862264, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1School of Command and Control Engineering, Army Engineering University of PLA, Nanjing 210007), AuthorCompanyExt(id=1329420565737222394, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, companyId=1329420565707862264, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1中国人民解放军陆军工程大学指挥控制工程学院 南京 210007)])]), Author(id=1329420566831935761, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, orderNo=3, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329420566911627539, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, authorId=1329420566831935761, language=EN, stringName=Zejun Chen, firstName=Zejun, middleName=null, lastName=Chen, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1School of Command and Control Engineering, Army Engineering University of PLA, Nanjing 210007, bio={"content":"

born in 1998. Master candidate. Her main research interests include intelligent education and exercise recommendation

"}, bioImg=null, bioContent=

born in 1998. Master candidate. Her main research interests include intelligent education and exercise recommendation

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329420566978736404, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, authorId=1329420566831935761, language=CN, stringName=陈泽君, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1中国人民解放军陆军工程大学指挥控制工程学院 南京 210007, bio={"img":"P02YoF05srLbDS6r8PXiqQ==","content":"

1998年生。硕士研究生。主要研究方向为智能教育、学习推荐

"}, bioImg=P02YoF05srLbDS6r8PXiqQ==, bioContent=

1998年生。硕士研究生。主要研究方向为智能教育、学习推荐

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329420565707862264, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, xref=1, ext=[AuthorCompanyExt(id=1329420565728833785, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, companyId=1329420565707862264, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1School of Command and Control Engineering, Army Engineering University of PLA, Nanjing 210007), AuthorCompanyExt(id=1329420565737222394, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, companyId=1329420565707862264, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1中国人民解放军陆军工程大学指挥控制工程学院 南京 210007)])]), Author(id=1329420568878756120, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, orderNo=4, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329420568958447900, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, authorId=1329420568878756120, language=EN, stringName=Xinjie Sun, firstName=Xinjie, middleName=null, lastName=Sun, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, 3, address=2School of Computer Science and Technology, University of Science and Technology of China, Hefei 230027
3School of Computer Science, Liupanshui Normal University, Liupanshui, Guizhou 553004, bio={"content":"

born in 1987. PhD, professor. His main research interests include intelligent education and adaptive learning

"}, bioImg=null, bioContent=

born in 1987. PhD, professor. His main research interests include intelligent education and adaptive learning

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329420569059111197, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, authorId=1329420568878756120, language=CN, stringName=孙新杰, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, 3, address=2中国科学技术大学计算机科学与技术学院 合肥 230027
3六盘水师范学院计算机科学学院 贵州六盘水 553004, bio={"img":"wscKhmLqaPHU57TebGXTdA==","content":"

1987年生。博士,教授。主要研究方向为智能教育、自适应学习

"}, bioImg=wscKhmLqaPHU57TebGXTdA==, bioContent=

1987年生。博士,教授。主要研究方向为智能教育、自适应学习

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329420565888217339, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, xref=2, ext=[AuthorCompanyExt(id=1329420565930160380, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, companyId=1329420565888217339, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2School of Computer Science and Technology, University of Science and Technology of China, Hefei 230027), AuthorCompanyExt(id=1329420565938548989, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, companyId=1329420565888217339, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2中国科学技术大学计算机科学与技术学院 合肥 230027)]), AuthorCompany(id=1329420566018240766, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, xref=3, ext=[AuthorCompanyExt(id=1329420566022435071, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, companyId=1329420566018240766, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3School of Computer Science, Liupanshui Normal University, Liupanshui, Guizhou 553004), AuthorCompanyExt(id=1329420566043406592, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, companyId=1329420566018240766, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3六盘水师范学院计算机科学学院 贵州六盘水 553004)])]), Author(id=1329420569130414367, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, orderNo=5, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329420569210106145, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, authorId=1329420569130414367, language=EN, stringName=Yifan Liu, firstName=Yifan, middleName=null, lastName=Liu, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1School of Command and Control Engineering, Army Engineering University of PLA, Nanjing 210007, bio={"content":"

born in 2000. Master candidate. His main research interests include intelligent education and data mining

"}, bioImg=null, bioContent=

born in 2000. Master candidate. His main research interests include intelligent education and data mining

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329420569293992226, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, authorId=1329420569130414367, language=CN, stringName=刘逸凡, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1中国人民解放军陆军工程大学指挥控制工程学院 南京 210007, bio={"img":"OSUvsHNf65S+I0kTarvEFg==","content":"

2000年生。硕士研究生。主要研究方向为智能教育、数据挖掘

"}, bioImg=OSUvsHNf65S+I0kTarvEFg==, bioContent=

2000年生。硕士研究生。主要研究方向为智能教育、数据挖掘

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329420565707862264, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, xref=1, ext=[AuthorCompanyExt(id=1329420565728833785, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, companyId=1329420565707862264, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1School of Command and Control Engineering, Army Engineering University of PLA, Nanjing 210007), AuthorCompanyExt(id=1329420565737222394, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, companyId=1329420565707862264, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1中国人民解放军陆军工程大学指挥控制工程学院 南京 210007)])]), Author(id=1329420569352712484, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, orderNo=6, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329420569453375782, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, authorId=1329420569352712484, language=EN, stringName=Enhong Chen, firstName=Enhong, middleName=null, lastName=Chen, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, address=2School of Computer Science and Technology, University of Science and Technology of China, Hefei 230027, bio={"content":"

born in 1968. PhD, professor. His main research interests include machine learning, data mining, education big data analysis, and personalized recommendation system

"}, bioImg=null, bioContent=

born in 1968. PhD, professor. His main research interests include machine learning, data mining, education big data analysis, and personalized recommendation system

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329420569516290343, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, authorId=1329420569352712484, language=CN, stringName=陈恩红, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, address=2中国科学技术大学计算机科学与技术学院 合肥 230027, bio={"img":"MED6poxD/PFxMJahToB+4A==","content":"

1968年生。博士,教授。主要研究方向为机器学习、数据挖掘、教育大数据分析、个性化推荐系统

"}, bioImg=MED6poxD/PFxMJahToB+4A==, bioContent=

1968年生。博士,教授。主要研究方向为机器学习、数据挖掘、教育大数据分析、个性化推荐系统

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329420565888217339, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, xref=2, ext=[AuthorCompanyExt(id=1329420565930160380, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, companyId=1329420565888217339, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2School of Computer Science and Technology, University of Science and Technology of China, Hefei 230027), AuthorCompanyExt(id=1329420565938548989, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, companyId=1329420565888217339, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2中国科学技术大学计算机科学与技术学院 合肥 230027)])])], keywords=[Keyword(id=1329420569600176424, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, language=EN, orderNo=1, keyword=knowledge tracing), Keyword(id=1329420569654702377, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, language=EN, orderNo=2, keyword=hybrid expert model), Keyword(id=1329420569713422634, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, language=EN, orderNo=3, keyword=agent reasoning), Keyword(id=1329420569784725803, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, language=EN, orderNo=4, keyword=contextual adaptation), Keyword(id=1329420569851834668, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, language=EN, orderNo=5, keyword=personalized learning), Keyword(id=1329420569914749229, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, language=EN, orderNo=6, keyword=multi-agent collaboration), Keyword(id=1329420569981858094, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, language=CN, orderNo=1, keyword=知识追踪), Keyword(id=1329420570048966959, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, language=CN, orderNo=2, keyword=混合专家模型), Keyword(id=1329420570111881520, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, language=CN, orderNo=3, keyword=智能体推理), Keyword(id=1329420570187378993, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, language=CN, orderNo=4, keyword=情境自适应), Keyword(id=1329420570288042290, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, language=CN, orderNo=5, keyword=个性化学习), Keyword(id=1329420570350956851, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, language=CN, orderNo=6, keyword=多智能体协同)], refs=[Reference(id=1329420574352322894, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=1, rfOrder=0, authorNames=null, journalName=null, refType=null, unstructuredReference=Abdelrahman G, Wang Qing, Nunes B. Knowledge tracing: A survey[J]. ACM Computing Surveys, 2023, 55(11): 1−37, articleTitle=null, refAbstract=null), Reference(id=1329420574427820367, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=2, rfOrder=1, authorNames=null, journalName=null, refType=null, unstructuredReference=刘铁园, 陈威, 常亮, 等. 基于深度学习的知识追踪研究进展[J]. 计算机研究与发展, 2022, 59(1): 81−104, articleTitle=null, refAbstract=null), Reference(id=1329420574494929232, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=2, rfOrder=2, authorNames=null, journalName=null, refType=null, unstructuredReference=Liu Tieyuan, Chen Wei, Chang Liang, et al. Research advances in the knowledge tracing based on deep learning[J]. Journal of Computer Research and Development, 2022, 59(1): 81−104 (in Chinese), articleTitle=null, refAbstract=null), Reference(id=1329420574557843794, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=3, rfOrder=3, authorNames=null, journalName=null, refType=null, unstructuredReference=Shen Shuanghong, Liu Qi, Huang Zhenya, et al. A survey of knowledge tracing: Models, variants, and applications[J]. IEEE Transactions on Learning Technologies, 2024, 17: 1858−1879, articleTitle=null, refAbstract=null), Reference(id=1329420574633341267, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=4, rfOrder=4, authorNames=null, journalName=null, refType=null, unstructuredReference=Corbett A T, Anderson J R. Knowledge tracing: Modeling the acquisition of procedural knowledge[J]. User Modeling and User-Adapted Interaction, 1994, 4(4): 253−278, articleTitle=null, refAbstract=null), Reference(id=1329420574700450132, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=5, rfOrder=5, authorNames=null, journalName=null, refType=null, unstructuredReference=Piech C, Bassen J, Huang J, et al. Deep knowledge tracing[J]. Advances in Neural Information Processing Systems, 2015, 28:1−12, articleTitle=null, refAbstract=null), Reference(id=1329420574771753301, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=6, rfOrder=6, authorNames=null, journalName=null, refType=null, unstructuredReference=Nakagawa H, Iwasawa Y, Matsuo Y. Graph-based knowledge tracing: Modeling student proficiency using graph neural network[C]//Proc of IEEE/WIC/ACM Int Conf on Web Intelligence. New York: ACM, 2019: 156−163, articleTitle=null, refAbstract=null), Reference(id=1329420574847250774, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=7, rfOrder=7, authorNames=null, journalName=null, refType=null, unstructuredReference=Pandey S, Karypis G. A self-attentive model for knowledge tracing[J]. arXiv preprint, arXiv: 1907.06837, 2019, articleTitle=null, refAbstract=null), Reference(id=1329420574918553943, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=8, rfOrder=8, authorNames=null, journalName=null, refType=null, unstructuredReference=Ghosh A, Heffernan N, Lan A S. Context-aware attentive knowledge tracing[C]//Proc of the 26th ACM SIGKDD Int Conf on Knowledge Discovery & Data Mining. New York: ACM, 2020: 2330−2339, articleTitle=null, refAbstract=null), Reference(id=1329420574994051416, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=9, rfOrder=9, authorNames=null, journalName=null, refType=null, unstructuredReference=Shen Shuanghong, Liu Qi, Chen Enhong, et al. Learning process-consistent knowledge tracing[C]//Proc of the 27th ACM SIGKDD Conf on Knowledge Discovery & Data Mining. New York: ACM, 2021: 1452−1460, articleTitle=null, refAbstract=null), Reference(id=1329420575065354585, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=10, rfOrder=10, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhu Siyi, Chen Wenjie. Deep learning based knowledge tracing: A review of the literature[C]//Proc of the 2025 Int Conf on Big Data and Informatization Education. New York: ACM, 2025: 81−87, articleTitle=null, refAbstract=null), Reference(id=1329420575145046362, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=11, rfOrder=11, authorNames=null, journalName=null, refType=null, unstructuredReference=Shen Shuanghong, Huang Zhenya, Liu Qi, et al. Assessing student’s dynamic knowledge state by exploring the question difficulty effect[C]//Proc of the 45th Int ACM SIGIR Conf on Research and Development in Information Retrieval. New York: ACM, 2022: 427−437, articleTitle=null, refAbstract=null), Reference(id=1329420575207960923, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=12, rfOrder=12, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhang Suojuan, Pu Jie, Cui Jing, et al. MLC-DKT: A multi-layer context-aware deep knowledge tracing model[J]. Knowledge-Based Systems, 2024, 303: 112384, articleTitle=null, refAbstract=null), Reference(id=1329420575270875484, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=13, rfOrder=13, authorNames=null, journalName=null, refType=null, unstructuredReference=Bai Yanhong, Zhao Jiabao, Wei Tingjiang, et al. A survey of explainable knowledge tracing[J]. arXiv preprint, arXiv: 2403.07279, 2024, articleTitle=null, refAbstract=null), Reference(id=1329420575354761565, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=14, rfOrder=14, authorNames=null, journalName=null, refType=null, unstructuredReference=Voultsiou E, Moussiades L. A systematic review of AI, VR, and LLM applications in special education: Opportunities, challenges, and future directions[J]. Education and Information Technologies, 2025, 30(13): 19141−19181, articleTitle=null, refAbstract=null), Reference(id=1329420575426064734, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=15, rfOrder=15, authorNames=null, journalName=null, refType=null, unstructuredReference=孙新杰, 刘淇, 张凯, 等. 2T-Agent: 生成式智能体驱动的类教师思维认知诊断框架[J]. 计算机学报, 2026, 49(3): 574−590, articleTitle=null, refAbstract=null), Reference(id=1329420575497367903, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=15, rfOrder=16, authorNames=null, journalName=null, refType=null, unstructuredReference=Sun Xinjie, Liu Qi, Zhang Kai, et al. 2T-Agent: A generative agent-driven cognitive diagnostic framework for teacher-like thinking[J]. Chinese Journal of Computers, 2026, 49(3): 574−590 (in Chinese), articleTitle=null, refAbstract=null), Reference(id=1329420575560282464, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=16, rfOrder=17, authorNames=null, journalName=null, refType=null, unstructuredReference=Wang Jian, Dai Yinpei, Zhang Yichi, et al. Training turn-by-turn verifiers for dialogue tutoring agents: The curious case of LLMs as your coding tutors[C]//Findings of the Association for Computational Linguistics: ACL 2025. Stroudsburg, PA: ACL, 2025: 12416−12436, articleTitle=null, refAbstract=null), Reference(id=1329420575639974241, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=17, rfOrder=18, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhang Jiani, Shi Xingjian, King I, et al. Dynamic key-value memory networks for knowledge tracing[C]//Proc of the 26th Int Conf on World Wide Web. New York: ACM, 2017: 765−774, articleTitle=null, refAbstract=null), Reference(id=1329420575719666018, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=18, rfOrder=19, authorNames=null, journalName=null, refType=null, unstructuredReference=Li Runze, Wu Siyu, Wang Jun, et al. CIKT: A collaborative and iterative knowledge tracing framework with large language models[J]. arXiv preprint, arXiv: 2505.17705, 2025, articleTitle=null, refAbstract=null), Reference(id=1329420575790969187, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=19, rfOrder=20, authorNames=null, journalName=null, refType=null, unstructuredReference=Fu Lingyue, Guan Hao, Du Kounianhua, et al. Sinkt: A structure-aware inductive knowledge tracing model with large language model[C]//Proc of the 33rd ACM Int Conf on Information and Knowledge Management. New York: ACM, 2024: 632−642, articleTitle=null, refAbstract=null), Reference(id=1329420575870660964, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=20, rfOrder=21, authorNames=null, journalName=null, refType=null, unstructuredReference=Wang Ziwei, Zhou Jie, Chen Qin, et al. LLM-KT: Aligning large language models with knowledge tracing using a plug-and-play instruction[J]. arXiv preprint, arXiv: 2502.02945, 2025, articleTitle=null, refAbstract=null), Reference(id=1329420575954547045, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=21, rfOrder=22, authorNames=null, journalName=null, refType=null, unstructuredReference=Brown J S, Collins A, Duguid P. Situated cognition and the culture of learning[J]. Educational Researcher, 1989, 18(1): 32−42, articleTitle=null, refAbstract=null), Reference(id=1329420576013267302, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=22, rfOrder=23, authorNames=null, journalName=null, refType=null, unstructuredReference=Lave J, Wenger E. Situated Learning: Legitimate Peripheral Participation[M]. Cambridge, UK: Cambridge University Press, 1991, articleTitle=null, refAbstract=null), Reference(id=1329420576076181863, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=23, rfOrder=24, authorNames=null, journalName=null, refType=null, unstructuredReference=蒲杰, 张所娟, 陈卫卫. 知识情境感知的深度知识追踪模型[J]. 南京大学学报 (自然科学版), 2024, 60(1): 76−86, articleTitle=null, refAbstract=null), Reference(id=1329420576139096424, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=23, rfOrder=25, authorNames=null, journalName=null, refType=null, unstructuredReference=Pu Jie, Zhang Suojuan, Chen Weiwei. A deep knowledge tracking model for contextual awareness of knowledge[J]. Journal of Nanjing University (Natural Sciences Edition), 2024, 60(1): 76−86 (in Chinese), articleTitle=null, refAbstract=null), Reference(id=1329420576197816681, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=24, rfOrder=26, authorNames=null, journalName=null, refType=null, unstructuredReference=Cheng Yan, Wu Gang. Zou Haifeng, et al. A knowledge query network model based on rasch model embedding for personalized online learning[J]. Frontiers in Psychology, 2022, 13: 846621, articleTitle=null, refAbstract=null), Reference(id=1329420576277508458, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=25, rfOrder=27, authorNames=null, journalName=null, refType=null, unstructuredReference=Yu Minghe, Li Fan, Liu Hengyu, et al. ContextKT: A context-based method for knowledge tracing[J]. Applied Sciences, 2022, 12(17): 8822, articleTitle=null, refAbstract=null), Reference(id=1329420576332034411, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=26, rfOrder=28, authorNames=null, journalName=null, refType=null, unstructuredReference=Yamkovenko B, Hogg C A R, Miller-Vedam M, et al. Practical evaluation of deep knowledge tracing models for use in learning platforms[J/OL]. International Educational Data Mining Society, 2025[2026-05-29]. https://educationaldatamining.org/EDM2025/Proc/2025.EDM.industry-papers.46/, articleTitle=null, refAbstract=null), Reference(id=1329420576407531884, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=27, rfOrder=29, authorNames=null, journalName=null, refType=null, unstructuredReference=Jacobs R A, Jordan M I, Nowlan S J, et al. Adaptive mixtures of local experts[J]. Neural Computation, 1991, 3(1): 79−87, articleTitle=null, refAbstract=null), Reference(id=1329420576470446445, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=28, rfOrder=30, authorNames=null, journalName=null, refType=null, unstructuredReference=Fedus W, Zoph B, Shazeer N. Switch Transformers: Scaling to trillion parameter models with simple and efficient sparsity[J]. Journal of Machine Learning Research, 2022, 23(120): 1−39, articleTitle=null, refAbstract=null), Reference(id=1329420576537555310, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=29, rfOrder=31, authorNames=null, journalName=null, refType=null, unstructuredReference=Jiang A Q, Sablayrolles A, Roux A, et al. Mixtral of experts[J]. arXiv preprint, arXiv: 2401.04088, 2024, articleTitle=null, refAbstract=null), Reference(id=1329420576608858479, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=30, rfOrder=32, authorNames=null, journalName=null, refType=null, unstructuredReference=He X O. Mixture of a million experts[J]. arXiv preprint, arXiv: 2407.04153, 2024, articleTitle=null, refAbstract=null), Reference(id=1329420576675967344, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=31, rfOrder=33, authorNames=null, journalName=null, refType=null, unstructuredReference=Liao Han, Zu Shuaishuai. RouterKT: Mixture-of-experts for knowledge tracing[J]. arXiv preprint, arXiv: 2504.08989, 2025, articleTitle=null, refAbstract=null), Reference(id=1329420576772436337, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=32, rfOrder=34, authorNames=null, journalName=null, refType=null, unstructuredReference=Nikolova E. Artificial intelligence tools into higher mathematics education: Opportunities, challenges, and student perceptions[J]. Mathematics and Informatics, 2024, 67(4): 406−417, articleTitle=null, refAbstract=null), Reference(id=1329420576826962290, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=33, rfOrder=35, authorNames=null, journalName=null, refType=null, unstructuredReference=陈慧敏, 刘知远, 孙茂松. 大语言模型时代的社会机遇与挑战[J]. 计算机研究与发展, 2024, 61(5): 1094−1103, articleTitle=null, refAbstract=null), Reference(id=1329420576885682547, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=33, rfOrder=36, authorNames=null, journalName=null, refType=null, unstructuredReference=Chen Huimin, Liu Zhiyuan, Sun Maosong. The social opportunities and challenges in the era of large language models[J]. Journal of Computer Research and Development, 2024, 61(5): 1094−1103 (in Chinese), articleTitle=null, refAbstract=null), Reference(id=1329420576948597108, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=34, rfOrder=37, authorNames=null, journalName=null, refType=null, unstructuredReference=Li Haoxuan, Yu Jifan, Ouyang Yuanxin, et al. Explainable few-shot knowledge tracing[J]. Frontiers of Digital Education, 2025, 2(4): 34, articleTitle=null, refAbstract=null), Reference(id=1329420577032483189, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=35, rfOrder=38, authorNames=null, journalName=null, refType=null, unstructuredReference=Berthon A, van der Schaar M. Language bottleneck models: A framework for interpretable knowledge tracing and beyond[J]. arXiv preprint, arXiv: 2506.16982, 2025, articleTitle=null, refAbstract=null), Reference(id=1329420577095397750, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=36, rfOrder=39, authorNames=null, journalName=null, refType=null, unstructuredReference=Gao Weibo, Liu Qi, Yue Linan, et al. Agent4edu: Generating learner response data by generative agents for intelligent education systems[C]//Proc of the AAAI Conf on Artificial Intelligence. Palo Alto, CA: AAAI, 2025, 39(22): 23923−23932, articleTitle=null, refAbstract=null), Reference(id=1329420577162506615, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=37, rfOrder=40, authorNames=null, journalName=null, refType=null, unstructuredReference=高文. 情境学习与情境认知[J]. 教育发展研究, 2001(8): 30−35, articleTitle=null, refAbstract=null), Reference(id=1329420577229615480, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=37, rfOrder=41, authorNames=null, journalName=null, refType=null, unstructuredReference=Gao Wen. Situational learning and situational cognition[J]. Educational Development Research, 2001(8): 30−35 (in Chinese), articleTitle=null, refAbstract=null), Reference(id=1329420577288335737, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=38, rfOrder=42, authorNames=null, journalName=null, refType=null, unstructuredReference=Lord F M. Applications of Item Response Theory to Practical Testing Problems[M]. Abingdon-on-Thames, UK: Routledge, 2012, articleTitle=null, refAbstract=null), Reference(id=1329420577351250298, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=39, rfOrder=43, authorNames=null, journalName=null, refType=null, unstructuredReference=Sweller J. Cognitive load theory, learning difficulty, and instructional design[J]. Learning and Instruction, 1994, 4(4): 295−312, articleTitle=null, refAbstract=null), Reference(id=1329420577422553467, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=40, rfOrder=44, authorNames=null, journalName=null, refType=null, unstructuredReference=Ausubel D P, Novak J D, Hanesian H. Educational Psychology: A Cognitive View[M]. New York: Holt, Rinehart and Winston, 1978, articleTitle=null, refAbstract=null), Reference(id=1329420577506439548, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=41, rfOrder=45, authorNames=null, journalName=null, refType=null, unstructuredReference=Ghosh A, Raspat J, Lan A. Option tracing: Beyond correctness analysis in knowledge tracing[C]//Proc of Int Conf on Artificial Intelligence in Education. Cham, Switzerland: Springer, 2021: 137−149, articleTitle=null, refAbstract=null), Reference(id=1329420577573548413, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=42, rfOrder=46, authorNames=null, journalName=null, refType=null, unstructuredReference=Karpicke J D, Roediger III H L. The critical importance of retrieval for learning[J]. Science, 2008, 319(5865): 966−968, articleTitle=null, refAbstract=null), Reference(id=1329420577632268670, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=43, rfOrder=47, authorNames=null, journalName=null, refType=null, unstructuredReference=Nagatani K, Zhang Qian, Sato M, et al. Augmenting knowledge tracing by considering forgetting behavior[C]//Proc of the World Wide Web Conf. New York: ACM, 2019: 3101−3107, articleTitle=null, refAbstract=null), Reference(id=1329420577699377535, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=44, rfOrder=48, authorNames=null, journalName=null, refType=null, unstructuredReference=Tulving E. Episodic and semantic memory[J]. Organization of Memory, 1972, 1(1): 381−403, articleTitle=null, refAbstract=null), Reference(id=1329420577758097792, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=45, rfOrder=49, authorNames=null, journalName=null, refType=null, unstructuredReference=Park J S, O'Brien J C, Cai C J, et al. Generative agents: Interactive simulacra of human behavior[C]//Proc of the 36th Annual ACM Symp on User Interface Software and Technology. New York: ACM, 2023: 1−22, articleTitle=null, refAbstract=null), Reference(id=1329420577816818049, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=46, rfOrder=50, authorNames=null, journalName=null, refType=null, unstructuredReference=隋洁, 吴艳红. 心理时间之旅——情景记忆的独特性[J]. 北京大学学报(自然科学版), 2004(2): 326−332, articleTitle=null, refAbstract=null), Reference(id=1329420577879732610, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=46, rfOrder=51, authorNames=null, journalName=null, refType=null, unstructuredReference=Sui Jie, Wu Yanhong. A psychological journey through time: The uniqueness of situational memory[J]. Journal of Peking University (Natural Sciences Edition), 2004(2): 326−332 (in Chinese), articleTitle=null, refAbstract=null), Reference(id=1329420577942647171, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=47, rfOrder=52, authorNames=null, journalName=null, refType=null, unstructuredReference=Anderson J R. Acquisition of cognitive skill[J]. Psychological Review, 1982, 89(4): 369−406, articleTitle=null, refAbstract=null), Reference(id=1329420578013950340, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=48, rfOrder=53, authorNames=null, journalName=null, refType=null, unstructuredReference=许嘉, 唐嵘蓉, 吕品, 等. 基于学习迁移的稳定知识追踪模型[J]. 华南师范大学学报(自然科学版), 2024, 56(4): 68−79, articleTitle=null, refAbstract=null), Reference(id=1329420578093642117, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=48, rfOrder=54, authorNames=null, journalName=null, refType=null, unstructuredReference=Xu Jia, Tang Rongrong, Lü Pin, et al. A stable knowledge tracking model based on learning transfer[J]. Journal of South China Normal University (Natural Sciences Edition), 2024, 56(4): 68−79 (in Chinese), articleTitle=null, refAbstract=null), Reference(id=1329420578173333894, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=49, rfOrder=55, authorNames=null, journalName=null, refType=null, unstructuredReference=Wang Qingsong, Yao Chang, Chen Jingyuan. Causal-Steer: Disentangled continuous style control without parallel corpora[C/OL]//Proc of the 14th Int Conf on Learning Representations. [2026-05-29]. https://openreview.net/pdf?id=yfiOYXvsX5, articleTitle=null, refAbstract=null), Reference(id=1329420578253025671, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=50, rfOrder=56, authorNames=null, journalName=null, refType=null, unstructuredReference=Alhafni B, Kulkarni V, Kumar D, et al. Personalized text generation with fine-grained linguistic control[C]//Proc of the 1st Workshop on Personalization of Generative AI Systems (PERSONALIZE 2024). Stroudsburg, PA: ACL, 2024: 88−101, articleTitle=null, refAbstract=null), Reference(id=1329420578315940232, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=51, rfOrder=57, authorNames=null, journalName=null, refType=null, unstructuredReference=Liu Zitao, Liu Qiongqiong, Chen Jiahao, et al. pyKT: A Python library to benchmark deep learning based knowledge tracing models[J]. Advances in Neural Information Processing Systems, 2022, 35: 18542−18555, articleTitle=null, refAbstract=null), Reference(id=1329420578378854793, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=52, rfOrder=58, authorNames=null, journalName=null, refType=null, unstructuredReference=Guo Teng, Qin Yu, Xia Yubin, et al. Enhancing knowledge tracing through decoupling cognitive pattern from error-prone data[C]//Proc of the ACM on Web Conf 2025. New York: ACM, 2025: 5108−5116, articleTitle=null, refAbstract=null), Reference(id=1329420578450157962, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=53, rfOrder=59, authorNames=null, journalName=null, refType=null, unstructuredReference=Bai Youheng, Li Xueyi, Liu Zitao, et al. Rethinking and improving student learning and forgetting processes for attention based knowledge tracing models[C]//Proc of the AAAI Conf on Artificial Intelligence. Palo Alto, CA: AAAI, 2025, 39(27): 27822−27830, articleTitle=null, refAbstract=null), Reference(id=1329420578508878219, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=54, rfOrder=60, authorNames=null, journalName=null, refType=null, unstructuredReference=Cheng Weihua, Du Hanwen, Li Chunxiao, et al. Uncertainty-aware knowledge tracing[C]//Proc of the AAAI Conf on Artificial Intelligence. Palo Alto, CA: AAAI, 2025, 39(27): 27905−27913, articleTitle=null, refAbstract=null)], funds=null, companyList=[AuthorCompany(id=1329420565707862264, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, xref=1, ext=[AuthorCompanyExt(id=1329420565728833785, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, companyId=1329420565707862264, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1School of Command and Control Engineering, Army Engineering University of PLA, Nanjing 210007), AuthorCompanyExt(id=1329420565737222394, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, companyId=1329420565707862264, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1中国人民解放军陆军工程大学指挥控制工程学院 南京 210007)]), AuthorCompany(id=1329420565888217339, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, xref=2, ext=[AuthorCompanyExt(id=1329420565930160380, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, companyId=1329420565888217339, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2School of Computer Science and Technology, University of Science and Technology of China, Hefei 230027), AuthorCompanyExt(id=1329420565938548989, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, companyId=1329420565888217339, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2中国科学技术大学计算机科学与技术学院 合肥 230027)]), AuthorCompany(id=1329420566018240766, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, xref=3, ext=[AuthorCompanyExt(id=1329420566022435071, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, companyId=1329420566018240766, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3School of Computer Science, Liupanshui Normal University, Liupanshui, Guizhou 553004), AuthorCompanyExt(id=1329420566043406592, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, companyId=1329420566018240766, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3六盘水师范学院计算机科学学院 贵州六盘水 553004)])], figs=[ArticleFig(id=1329420570480980276, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, language=EN, label=Figure 1, caption=Example of context-aware knowledge tracing, figureFileSmall=Qr39ePJfrZlqXvnCumx5HQ==, figureFileBig=S0UuvqUXv2p7HXu+3PuXsQ==, tableContent=null), ArticleFig(id=1329420570556477749, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, language=CN, label=图1, caption=基于情境感知的知识追踪示例, figureFileSmall=Qr39ePJfrZlqXvnCumx5HQ==, figureFileBig=S0UuvqUXv2p7HXu+3PuXsQ==, tableContent=null), ArticleFig(id=1329420570795553078, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, language=EN, label=Figure 2, caption=CAKT-Agent framework, figureFileSmall=TzrxxCW/oXlSeEy7T3A3Dw==, figureFileBig=/wG3bKE04LIRYrdq4pA8Pg==, tableContent=null), ArticleFig(id=1329420570862661943, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, language=CN, label=图2, caption=CAKT-Agent框架, figureFileSmall=TzrxxCW/oXlSeEy7T3A3Dw==, figureFileBig=/wG3bKE04LIRYrdq4pA8Pg==, tableContent=null), ArticleFig(id=1329420570921382200, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, language=EN, label=Figure 3, caption=Performance of the cumulative multi-step forecasting task, figureFileSmall=F0KmrGdnNBEjO3JLgkYuFA==, figureFileBig=t/fPwq+fHnORCm8OEmUXVw==, tableContent=null), ArticleFig(id=1329420570992685369, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, language=CN, label=图3, caption=累积式多步预测任务表现, figureFileSmall=F0KmrGdnNBEjO3JLgkYuFA==, figureFileBig=t/fPwq+fHnORCm8OEmUXVw==, tableContent=null), ArticleFig(id=1329420571097542970, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, language=EN, label=Figure 4, caption=Ablation experimental results, figureFileSmall=9lJFYvxqbvmYYdwo14GslQ==, figureFileBig=A9P17Bbvn/9UviF2kgPYnA==, tableContent=null), ArticleFig(id=1329420571164651835, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, language=CN, label=图4, caption=消融实验结果, figureFileSmall=9lJFYvxqbvmYYdwo14GslQ==, figureFileBig=A9P17Bbvn/9UviF2kgPYnA==, tableContent=null), ArticleFig(id=1329420573077254460, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, language=EN, label=Figure 5, caption=Sensitivity test results, figureFileSmall=/hTd4RhaYnhZWoAG/+CdnQ==, figureFileBig=2ckp8IO17+G8rHxGZdUTOw==, tableContent=null), ArticleFig(id=1329420573148557629, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, language=CN, label=图5, caption=敏感性实验结果, figureFileSmall=/hTd4RhaYnhZWoAG/+CdnQ==, figureFileBig=2ckp8IO17+G8rHxGZdUTOw==, tableContent=null), ArticleFig(id=1329420573219860798, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, language=EN, label=Figure 6, caption=Results of the LLM fine-tuning capability assessmental experiment, figureFileSmall=I9AXrrpaDV1j+zwIiMd3mQ==, figureFileBig=7OH7JLvq+qc+3gWMf9wEGQ==, tableContent=null), ArticleFig(id=1329420573282775359, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, language=CN, label=图6, caption=LLM微调能力评估实验结果, figureFileSmall=I9AXrrpaDV1j+zwIiMd3mQ==, figureFileBig=7OH7JLvq+qc+3gWMf9wEGQ==, tableContent=null), ArticleFig(id=1329420573362467136, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, language=EN, label=Figure 7, caption=Case study, figureFileSmall=w/1sPRiWVpXMOier199ZuA==, figureFileBig=gK6MFJtemnuyTO7b3MmHoA==, tableContent=null), ArticleFig(id=1329420573429576001, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, language=CN, label=图7, caption=案例分析, figureFileSmall=w/1sPRiWVpXMOier199ZuA==, figureFileBig=gK6MFJtemnuyTO7b3MmHoA==, tableContent=null), ArticleFig(id=1329420573492490562, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, language=EN, label=Figure 8, caption=Expert score comparison results, figureFileSmall=WtbZT22sD7JQsUi4heDdjw==, figureFileBig=CT19CBLZT0+TLAp0AtqUsg==, tableContent=null), ArticleFig(id=1329420573555405123, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, language=CN, label=图8, caption=专家评分对比结果, figureFileSmall=WtbZT22sD7JQsUi4heDdjw==, figureFileBig=CT19CBLZT0+TLAp0AtqUsg==, tableContent=null), ArticleFig(id=1329420573622513988, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, language=EN, label=Table 1, caption=

Dataset Statistics

, figureFileSmall=null, figureFileBig=null, tableContent=
参数NIPS34ASSIST2012
交互数量ins13819582541201
学生数量s491819368
问题数量q94845716
知识概念数量c57265
ins/s281131.2
c/q1.011
), ArticleFig(id=1329420573698011461, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, language=CN, label=表1, caption=

数据集统计信息

, figureFileSmall=null, figureFileBig=null, tableContent=
参数NIPS34ASSIST2012
交互数量ins13819582541201
学生数量s491819368
问题数量q94845716
知识概念数量c57265
ins/s281131.2
c/q1.011
), ArticleFig(id=1329420573756731718, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, language=EN, label=Table 2, caption=

Performance Comparison of Knowledge Tracing Models on Single-Step Forecasting Tasks %

, figureFileSmall=null, figureFileBig=null, tableContent=
模型NIPS34ASSIST2012
AUC↑ACC↑RMSE↓AUC↑ACC↑RMSE↓
DKT77.10±0.059470.44±0.061243.90±0.025173.07±0.018073.37±0.016142.45±0.0072
DKVMN76.88±0.080270.21±0.065244.00±0.041672.92±0.012073.42±0.025542.47±0.0105
AKT80.50±0.030073.32±0.028942.25±0.031177.58±0.280775.55±0.231540.96±0.1359
RobustKT80.43±0.020873.21±0.058342.32±0.030377.88±0.049475.58±0.080840.85±0.0315
LPKT79.49±0.015072.53±0.046142.75±0.010477.26±0.013275.41±0.017840.97±0.0061
DIMKT80.39±0.011073.20±0.020042.26±0.021076.86±0.238475.11±0.147441.15±0.0954
LefoKT80.08±0.029473.04±0.038542.42±0.010677.54±0.208475.49±0.155540.99±0.1215
UKT79.87±0.076172.73±0.096042.55±0.054077.30±0.026575.13±0.040041.15±0.0195
CAKT(本文)87.61±0.068979.47±0.055937.66±0.043779.87±0.320976.82±0.254539.96±0.1438
), ArticleFig(id=1329420573836423495, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, language=CN, label=表2, caption=

知识追踪模型在单步预测任务上的性能对比

, figureFileSmall=null, figureFileBig=null, tableContent=
模型NIPS34ASSIST2012
AUC↑ACC↑RMSE↓AUC↑ACC↑RMSE↓
DKT77.10±0.059470.44±0.061243.90±0.025173.07±0.018073.37±0.016142.45±0.0072
DKVMN76.88±0.080270.21±0.065244.00±0.041672.92±0.012073.42±0.025542.47±0.0105
AKT80.50±0.030073.32±0.028942.25±0.031177.58±0.280775.55±0.231540.96±0.1359
RobustKT80.43±0.020873.21±0.058342.32±0.030377.88±0.049475.58±0.080840.85±0.0315
LPKT79.49±0.015072.53±0.046142.75±0.010477.26±0.013275.41±0.017840.97±0.0061
DIMKT80.39±0.011073.20±0.020042.26±0.021076.86±0.238475.11±0.147441.15±0.0954
LefoKT80.08±0.029473.04±0.038542.42±0.010677.54±0.208475.49±0.155540.99±0.1215
UKT79.87±0.076172.73±0.096042.55±0.054077.30±0.026575.13±0.040041.15±0.0195
CAKT(本文)87.61±0.068979.47±0.055937.66±0.043779.87±0.320976.82±0.254539.96±0.1438
), ArticleFig(id=1329420573911920968, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, language=EN, label=Table 3, caption=

Performance Comparison of Knowledge Tracing Models on Non-Cumulative Multi-Step Forecasting Tasks %

, figureFileSmall=null, figureFileBig=null, tableContent=
模型NIPS34ASSIST2012
AUC↑ACC↑RMSE↓AUC↑ACC↑RMSE↓
DKT77.30±0.065570.62±0.054943.80±0.024273.20±0.003373.42±0.026042.41±0.0090
DKVMN77.14±0.087670.44±0.085643.88±0.041673.08±0.015973.49±0.025042.42±0.0087
AKT80.76±0.029173.53±0.039142.08±0.027777.74±0.308275.63±0.231540.89±0.1462
RobustKT80.65±0.027073.39±0.063042.19±0.031478.04±0.045275.65±0.090740.79±0.0371
LPKT79.75±0.021972.61±0.034642.68±0.018177.42±0.007975.49±0.035940.91±0.0180
DIMKT80.60±0.012873.40±0.006142.15±0.023376.98±0.235875.16±0.155341.11±0.0948
LefoKT80.25±0.028273.18±0.048342.32±0.013577.70±0.208375.56±0.149040.92±0.1043
UKT80.00±0.074772.83±0.083342.48±0.047477.44±0.029175.17±0.041141.11±0.0292
CAKT(本文)81.93±0.166674.27±0.123741.66±0.109478.49±0.222675.94±0.227040.62±0.0983
), ArticleFig(id=1329420573987418441, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, language=CN, label=表3, caption=

知识追踪模型在非累积式多步预测任务上的性能对比

, figureFileSmall=null, figureFileBig=null, tableContent=
模型NIPS34ASSIST2012
AUC↑ACC↑RMSE↓AUC↑ACC↑RMSE↓
DKT77.30±0.065570.62±0.054943.80±0.024273.20±0.003373.42±0.026042.41±0.0090
DKVMN77.14±0.087670.44±0.085643.88±0.041673.08±0.015973.49±0.025042.42±0.0087
AKT80.76±0.029173.53±0.039142.08±0.027777.74±0.308275.63±0.231540.89±0.1462
RobustKT80.65±0.027073.39±0.063042.19±0.031478.04±0.045275.65±0.090740.79±0.0371
LPKT79.75±0.021972.61±0.034642.68±0.018177.42±0.007975.49±0.035940.91±0.0180
DIMKT80.60±0.012873.40±0.006142.15±0.023376.98±0.235875.16±0.155341.11±0.0948
LefoKT80.25±0.028273.18±0.048342.32±0.013577.70±0.208375.56±0.149040.92±0.1043
UKT80.00±0.074772.83±0.083342.48±0.047477.44±0.029175.17±0.041141.11±0.0292
CAKT(本文)81.93±0.166674.27±0.123741.66±0.109478.49±0.222675.94±0.227040.62±0.0983
), ArticleFig(id=1329420574054527306, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, language=EN, label=Table 4, caption=

Efficiency Comparison on NIPS34 Dataset

, figureFileSmall=null, figureFileBig=null, tableContent=
模型离线训练阶段在线服务阶段
单轮epoch
耗时/s
参数量Checkpoint/
MB
GPU 峰值
显存/MB
每个请求的
平均延迟/ms
GPU 峰值
显存/MB
参数量端到端
延迟/s
单次调用
Token 总量
DKT0.860.36×1061.36724.791.5855.200.36×10614.20758
DKVMN2.450.22×1060.822579.336.8260.360.22×10614.21758
AKT16.877.04×10626.958944.6716.4880.177.04×10614.21758
RobustKT4.186.32×10624.188921.0115.8277.416.32×10614.21758
LPKT20.201.49×1065.681075.1683.8939.301.49×10614.28758
DIMKT8.210.66×1062.55353.4737.1937.430.66×10614.23758
LefoKT8.191.84×1067.035783.613.6656.641.84×10614.20758
UKT7.674.43×10616.9618160.387.4672.374.43×10614.21758
CAKT-LLM(本文)11.061.55×1066.021138.5850.4447.921.55×10614.25864
), ArticleFig(id=1329420574117441867, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, language=CN, label=表4, caption=

在NIPS34数据集上的效率对比

, figureFileSmall=null, figureFileBig=null, tableContent=
模型离线训练阶段在线服务阶段
单轮epoch
耗时/s
参数量Checkpoint/
MB
GPU 峰值
显存/MB
每个请求的
平均延迟/ms
GPU 峰值
显存/MB
参数量端到端
延迟/s
单次调用
Token 总量
DKT0.860.36×1061.36724.791.5855.200.36×10614.20758
DKVMN2.450.22×1060.822579.336.8260.360.22×10614.21758
AKT16.877.04×10626.958944.6716.4880.177.04×10614.21758
RobustKT4.186.32×10624.188921.0115.8277.416.32×10614.21758
LPKT20.201.49×1065.681075.1683.8939.301.49×10614.28758
DIMKT8.210.66×1062.55353.4737.1937.430.66×10614.23758
LefoKT8.191.84×1067.035783.613.6656.641.84×10614.20758
UKT7.674.43×10616.9618160.387.4672.374.43×10614.21758
CAKT-LLM(本文)11.061.55×1066.021138.5850.4447.921.55×10614.25864
), ArticleFig(id=1329420574192939340, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, language=EN, label=Table 5, caption=

Artificial Diagnosis Consistency Analysis

, figureFileSmall=null, figureFileBig=null, tableContent=
评估维度PrecisionRecallF1ACCKappa
薄弱知识点0.8070.7610.784
错误模式0.9050.9500.927
教学干预策略0.8850.6000.715
学习状态0.6500.551
), ArticleFig(id=1329420574264242509, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, language=CN, label=表5, caption=

人工诊断一致性分析

, figureFileSmall=null, figureFileBig=null, tableContent=
评估维度PrecisionRecallF1ACCKappa
薄弱知识点0.8070.7610.784
错误模式0.9050.9500.927
教学干预策略0.8850.6000.715
学习状态0.6500.551
)], attaches=null, journal=Journal(id=1324759606947315908, delFlag=0, nameCn=计算机研究与发展, nameEn=Journal of Computer Research and Development, nameHistory1=null, nameHistory2=null, issn=1000-1239, eissn=null, cn=11-1777/TP, coden=null, periodic=0, language=CN, oaType=null, ccby=null, superviseOffice=null, ownerOffice=null, pubOffice=null, editorOffice=null, officeType=null, aims=null, clcCode=null, officeProv=null, officeCity=null, officeAddr=null, officeZip=null, officeEmail=null, officePhone=null, editDirector=null, officeDirector=null, officeDirectorPhone=null, officeStaffNum=null, officeEmpNum=null, coverPicUrl=k2l8zgqwstuxxevFX2qxVA==, journalPrice=null, startedYear=null, abbrevIsoEn=Journal of Computer Research and Development, journalRemark=null, publicationField=null, createdTime=1793776903141, updatedTime=1793850698612, createdBy=18614031015, updatedBy=13701087609, firstLetterCn=J, firstLetterEn=J, subjectCode=Electronic Information, subjectName=null, subjectCodeEn=Electronic Information, subjectNameEn=null, picCn=k2l8zgqwstuxxevFX2qxVA==, picEn=r6k0Ry+RD+c/VXj5QarmDw==, jcr=null, cjcr=null, exts=[JournalExt(id=1325069127710286688, language=CN, name=计算机研究与发展, nameHistory1=null, nameHistory2=null, managedBy=, sponsoredBy=, publishedBy=, editorOffice=, officeProv=null, officeCity=null, officeAddr=, officeZip=, editDirector=, officeDirector=null, officePhone=null, coverPicUrl=null, journalRemark=, submitArticleUrl=null, websiteUrl=, createdTime=1793850698640, updatedTime=1793850698640, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=, submissionAuthorUrl=https://crad.ict.ac.cn/journalx/authorLogOn.action, submissionEditorUrl=https://crad.ict.ac.cn/journalx/editorLogOn.action, submissionReviewUrl=https://crad.ict.ac.cn/journalx/expertLogOn.action, submissionCeEditorUrl=https://crad.ict.ac.cn/journalx/editorInChiefLogOn.action, submissionAeEditorUrl=, option={"copyright":""}), JournalExt(id=1325069127777395553, language=EN, name=Journal of Computer Research and Development, nameHistory1=null, nameHistory2=null, managedBy=, sponsoredBy=, publishedBy=, editorOffice=, officeProv=null, officeCity=null, officeAddr=, officeZip=, editDirector=, officeDirector=null, officePhone=null, coverPicUrl=null, journalRemark=, submitArticleUrl=null, websiteUrl=, createdTime=1793850698656, updatedTime=1793850698656, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=, submissionAuthorUrl=https://crad.ict.ac.cn/journalx/authorLogOn.action, submissionEditorUrl=https://crad.ict.ac.cn/journalx/editorLogOn.action, submissionReviewUrl=https://crad.ict.ac.cn/journalx/expertLogOn.action, submissionCeEditorUrl=https://crad.ict.ac.cn/journalx/editorInChiefLogOn.action, submissionAeEditorUrl=, option={"copyright":""})], databaseList=null, tenantJournalId=1324759914146529478, websiteList=[Website(id=1325070300139892771, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1324759914146529478, journalNameCn=null, journalNameEn=null, grayFlag=null, tenantId=1146029695717560320, platformId=null, journalGroupId=null, journalGroupNameCn=null, journalGroupNameEn=null, type=1, domain=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN, language=CN, createTime=1793850978169, createBy=18614031015, updateTime=1793851040259, updateBy=18614031015, name=计算机研究与发展-中文, tplId=1146099689490845704, title=计算机研究与发展, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1325070893751349360, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=articleTextType, value=kx, createTime=1793851119697, updateTime=1793851119697, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893730377837, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=banner, value=null, createTime=1793851119692, updateTime=1793851119692, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893772320883, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=grayFlag, value=0, createTime=1793851119702, updateTime=1793851119702, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893721989228, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=logo, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/file/pic?fileId=U27tDsy717YF7XSPA+Vwdw==, createTime=1793851119690, updateTime=1793851119690, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893784903797, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=minRunFlag, value=0, createTime=1793851119705, updateTime=1793851119705, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893742960751, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/file/pic, createTime=1793851119695, updateTime=1793851119695, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893780709492, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=silenceFlag, value=0, createTime=1793851119704, updateTime=1793851119704, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893738766446, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_cn_619/, createTime=1793851119694, updateTime=1793851119694, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893755543665, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=themeColor, value=null, createTime=1793851119698, updateTime=1793851119698, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893763932274, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=themeStyle, value=null, createTime=1793851119700, updateTime=1793851119700, creator=18614031015, updator=18614031015)]), Website(id=1325070300580294698, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1324759914146529478, journalNameCn=null, journalNameEn=null, grayFlag=null, tenantId=1146029695717560320, platformId=null, journalGroupId=null, journalGroupNameCn=null, journalGroupNameEn=null, type=1, domain=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN, language=EN, createTime=1793850978274, createBy=18614031015, updateTime=1793851036638, updateBy=18614031015, name=计算机研究与发展-英文, tplId=1146101810881728533, title=Journal of Computer Research and Development, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1325070925288321148, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=articleTextType, value=kx, createTime=1793851127216, updateTime=1793851127216, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925263155321, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=banner, value=null, createTime=1793851127210, updateTime=1793851127210, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925309292671, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=grayFlag, value=0, createTime=1793851127221, updateTime=1793851127221, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925254766712, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=logo, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/file/pic?fileId=U27tDsy717YF7XSPA+Vwdw==, createTime=1793851127208, updateTime=1793851127208, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925326069889, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=minRunFlag, value=0, createTime=1793851127225, updateTime=1793851127225, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925279932539, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/file/pic, createTime=1793851127214, updateTime=1793851127214, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925317681280, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=silenceFlag, value=0, createTime=1793851127223, updateTime=1793851127223, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925271543930, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_en_623/, createTime=1793851127212, updateTime=1793851127212, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925292515453, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=themeColor, value=null, createTime=1793851127217, updateTime=1793851127217, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925300904062, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=themeStyle, value=null, createTime=1793851127219, updateTime=1793851127219, creator=18614031015, updator=18614031015)])], journalTitle=计算机研究与发展, weixinUrl=null, journalUrl=https://crad.ict.ac.cn/, iacademicId=null, status=1, seqNo=null, journalTitleEn=Journal of Computer Research and Development, journalPhotoCn=k2l8zgqwstuxxevFX2qxVA==, journalPhotoEn=r6k0Ry+RD+c/VXj5QarmDw==, journalFirstLetter=J, journalRecommend=null, journalNew=null, journalCollection=null, jcrJf=null, cjcrJf=null, jcrJfStr=null, cjcrJfStr=null, submissionFirstDecision=null, sciSubjectClassification=null, casSubjectClassification=null, citeScore=null, totalCitationFrequency=null, icpCode=null, psCode=null, advertisingLicenseCode=null, copyrightInformation=null, country=null, option=, provinceCode=null, provinceName=null, collectFlag=false, interPubPlatform=, interPubPlatformUrl=null), detailUrlCn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/10.7544/issn1000-1239.202660421, detailUrlEn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/10.7544/issn1000-1239.202660421, pdfUrlCn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/PDF/10.7544/issn1000-1239.202660421, pdfUrlEn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/PDF/10.7544/issn1000-1239.202660421, aliStartDate=0, aliEndDate=0, collectionFlag=false, citedCount=null, citedUrl=null, previewStatus=0, delFlag=0, hasFullText=1, orderTime=1788192000000, fullTextJson=null, articleText=null, reference=null)
收藏切换
CAKT-Agent:基于MoE的情境自适应知识追踪智能体
龙佳琪 1 , 张所娟 1, * , 郝文宁 1 , 陈泽君 1 , 孙新杰 2, 3 , 刘逸凡 1 , 陈恩红 2
计算机研究与发展 | “知识增强的人工智能前沿技术及应用”专题 2026,63(9): 2319-2337
收起
收藏切换
计算机研究与发展 |“知识增强的人工智能前沿技术及应用”专题 2026 , 63 (9) : 2319 -2337
CAKT-Agent:基于MoE的情境自适应知识追踪智能体
全屏
[Author(id=1329420566156652802, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=jiaqilong@aeu.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329420566240538884, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, authorId=1329420566156652802, language=EN, stringName=Jiaqi Long, firstName=Jiaqi, middleName=null, lastName=Long, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1School of Command and Control Engineering, Army Engineering University of PLA, Nanjing 210007, bio={"content":"

born in 2002. Master candidate. Her main research interests include intelligent education and knowledge tracing

"}, bioImg=null, bioContent=

born in 2002. Master candidate. Her main research interests include intelligent education and knowledge tracing

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329420566316036357, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, authorId=1329420566156652802, language=CN, stringName=龙佳琪, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1中国人民解放军陆军工程大学指挥控制工程学院 南京 210007, bio={"img":"tYa6D1n98DG+b5+W73u8aA==","content":"

2002年生。硕士研究生。主要研究方向为智能教育、知识追踪

"}, bioImg=tYa6D1n98DG+b5+W73u8aA==, bioContent=

2002年生。硕士研究生。主要研究方向为智能教育、知识追踪

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329420565707862264, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, xref=1, ext=[AuthorCompanyExt(id=1329420565728833785, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, companyId=1329420565707862264, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1School of Command and Control Engineering, Army Engineering University of PLA, Nanjing 210007), AuthorCompanyExt(id=1329420565737222394, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, companyId=1329420565707862264, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1中国人民解放军陆军工程大学指挥控制工程学院 南京 210007)])]), Author(id=1329420566387339527, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, orderNo=1, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=suojuanzhang@aeu.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=1, authorType=1, ext={EN=AuthorExt(id=1329420566479614217, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, authorId=1329420566387339527, language=EN, stringName=Suojuan Zhang, firstName=Suojuan, middleName=null, lastName=Zhang, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, *, address=1School of Command and Control Engineering, Army Engineering University of PLA, Nanjing 210007, bio={"content":"

born in 1983. PhD, professor. Her main research interests include educational data mining, knowledge tracing, and personalized exercise recommendation

"}, bioImg=null, bioContent=

born in 1983. PhD, professor. Her main research interests include educational data mining, knowledge tracing, and personalized exercise recommendation

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329420566571888906, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, authorId=1329420566387339527, language=CN, stringName=张所娟, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, *, address=1中国人民解放军陆军工程大学指挥控制工程学院 南京 210007, bio={"img":"kp7oy3Hv4nf/jLmTywDAyg==","content":"

1983年生。博士,教授。主要研究方向为教育数据挖掘、知识追踪、个性化学习推荐

"}, bioImg=kp7oy3Hv4nf/jLmTywDAyg==, bioContent=

1983年生。博士,教授。主要研究方向为教育数据挖掘、知识追踪、个性化学习推荐

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329420565707862264, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, xref=1, ext=[AuthorCompanyExt(id=1329420565728833785, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, companyId=1329420565707862264, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1School of Command and Control Engineering, Army Engineering University of PLA, Nanjing 210007), AuthorCompanyExt(id=1329420565737222394, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, companyId=1329420565707862264, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1中国人民解放军陆军工程大学指挥控制工程学院 南京 210007)])]), Author(id=1329420566630609164, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, orderNo=2, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329420566701912334, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, authorId=1329420566630609164, language=EN, stringName=Wenning Hao, firstName=Wenning, middleName=null, lastName=Hao, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1School of Command and Control Engineering, Army Engineering University of PLA, Nanjing 210007, bio={"content":"

born in 1971. PhD, professor. His main research interests include large-scale high-dimensional data compression and evaluation of operational effectiveness

"}, bioImg=null, bioContent=

born in 1971. PhD, professor. His main research interests include large-scale high-dimensional data compression and evaluation of operational effectiveness

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329420566769021199, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, authorId=1329420566630609164, language=CN, stringName=郝文宁, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1中国人民解放军陆军工程大学指挥控制工程学院 南京 210007, bio={"img":"5ATAy/bWzMHW9yGN3NEsVA==","content":"

1971年生。博士,教授。主要研究方向为大规模高维数据压缩、作战效能评估

"}, bioImg=5ATAy/bWzMHW9yGN3NEsVA==, bioContent=

1971年生。博士,教授。主要研究方向为大规模高维数据压缩、作战效能评估

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329420565707862264, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, xref=1, ext=[AuthorCompanyExt(id=1329420565728833785, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, companyId=1329420565707862264, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1School of Command and Control Engineering, Army Engineering University of PLA, Nanjing 210007), AuthorCompanyExt(id=1329420565737222394, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, companyId=1329420565707862264, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1中国人民解放军陆军工程大学指挥控制工程学院 南京 210007)])]), Author(id=1329420566831935761, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, orderNo=3, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329420566911627539, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, authorId=1329420566831935761, language=EN, stringName=Zejun Chen, firstName=Zejun, middleName=null, lastName=Chen, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1School of Command and Control Engineering, Army Engineering University of PLA, Nanjing 210007, bio={"content":"

born in 1998. Master candidate. Her main research interests include intelligent education and exercise recommendation

"}, bioImg=null, bioContent=

born in 1998. Master candidate. Her main research interests include intelligent education and exercise recommendation

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329420566978736404, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, authorId=1329420566831935761, language=CN, stringName=陈泽君, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1中国人民解放军陆军工程大学指挥控制工程学院 南京 210007, bio={"img":"P02YoF05srLbDS6r8PXiqQ==","content":"

1998年生。硕士研究生。主要研究方向为智能教育、学习推荐

"}, bioImg=P02YoF05srLbDS6r8PXiqQ==, bioContent=

1998年生。硕士研究生。主要研究方向为智能教育、学习推荐

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329420565707862264, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, xref=1, ext=[AuthorCompanyExt(id=1329420565728833785, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, companyId=1329420565707862264, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1School of Command and Control Engineering, Army Engineering University of PLA, Nanjing 210007), AuthorCompanyExt(id=1329420565737222394, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, companyId=1329420565707862264, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1中国人民解放军陆军工程大学指挥控制工程学院 南京 210007)])]), Author(id=1329420568878756120, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, orderNo=4, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329420568958447900, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, authorId=1329420568878756120, language=EN, stringName=Xinjie Sun, firstName=Xinjie, middleName=null, lastName=Sun, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, 3, address=2School of Computer Science and Technology, University of Science and Technology of China, Hefei 230027
3School of Computer Science, Liupanshui Normal University, Liupanshui, Guizhou 553004, bio={"content":"

born in 1987. PhD, professor. His main research interests include intelligent education and adaptive learning

"}, bioImg=null, bioContent=

born in 1987. PhD, professor. His main research interests include intelligent education and adaptive learning

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329420569059111197, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, authorId=1329420568878756120, language=CN, stringName=孙新杰, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, 3, address=2中国科学技术大学计算机科学与技术学院 合肥 230027
3六盘水师范学院计算机科学学院 贵州六盘水 553004, bio={"img":"wscKhmLqaPHU57TebGXTdA==","content":"

1987年生。博士,教授。主要研究方向为智能教育、自适应学习

"}, bioImg=wscKhmLqaPHU57TebGXTdA==, bioContent=

1987年生。博士,教授。主要研究方向为智能教育、自适应学习

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329420565888217339, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, xref=2, ext=[AuthorCompanyExt(id=1329420565930160380, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, companyId=1329420565888217339, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2School of Computer Science and Technology, University of Science and Technology of China, Hefei 230027), AuthorCompanyExt(id=1329420565938548989, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, companyId=1329420565888217339, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2中国科学技术大学计算机科学与技术学院 合肥 230027)]), AuthorCompany(id=1329420566018240766, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, xref=3, ext=[AuthorCompanyExt(id=1329420566022435071, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, companyId=1329420566018240766, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3School of Computer Science, Liupanshui Normal University, Liupanshui, Guizhou 553004), AuthorCompanyExt(id=1329420566043406592, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, companyId=1329420566018240766, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3六盘水师范学院计算机科学学院 贵州六盘水 553004)])]), Author(id=1329420569130414367, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, orderNo=5, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329420569210106145, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, authorId=1329420569130414367, language=EN, stringName=Yifan Liu, firstName=Yifan, middleName=null, lastName=Liu, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1School of Command and Control Engineering, Army Engineering University of PLA, Nanjing 210007, bio={"content":"

born in 2000. Master candidate. His main research interests include intelligent education and data mining

"}, bioImg=null, bioContent=

born in 2000. Master candidate. His main research interests include intelligent education and data mining

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329420569293992226, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, authorId=1329420569130414367, language=CN, stringName=刘逸凡, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1中国人民解放军陆军工程大学指挥控制工程学院 南京 210007, bio={"img":"OSUvsHNf65S+I0kTarvEFg==","content":"

2000年生。硕士研究生。主要研究方向为智能教育、数据挖掘

"}, bioImg=OSUvsHNf65S+I0kTarvEFg==, bioContent=

2000年生。硕士研究生。主要研究方向为智能教育、数据挖掘

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329420565707862264, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, xref=1, ext=[AuthorCompanyExt(id=1329420565728833785, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, companyId=1329420565707862264, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1School of Command and Control Engineering, Army Engineering University of PLA, Nanjing 210007), AuthorCompanyExt(id=1329420565737222394, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, companyId=1329420565707862264, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1中国人民解放军陆军工程大学指挥控制工程学院 南京 210007)])]), Author(id=1329420569352712484, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, orderNo=6, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329420569453375782, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, authorId=1329420569352712484, language=EN, stringName=Enhong Chen, firstName=Enhong, middleName=null, lastName=Chen, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, address=2School of Computer Science and Technology, University of Science and Technology of China, Hefei 230027, bio={"content":"

born in 1968. PhD, professor. His main research interests include machine learning, data mining, education big data analysis, and personalized recommendation system

"}, bioImg=null, bioContent=

born in 1968. PhD, professor. His main research interests include machine learning, data mining, education big data analysis, and personalized recommendation system

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329420569516290343, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, authorId=1329420569352712484, language=CN, stringName=陈恩红, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, address=2中国科学技术大学计算机科学与技术学院 合肥 230027, bio={"img":"MED6poxD/PFxMJahToB+4A==","content":"

1968年生。博士,教授。主要研究方向为机器学习、数据挖掘、教育大数据分析、个性化推荐系统

"}, bioImg=MED6poxD/PFxMJahToB+4A==, bioContent=

1968年生。博士,教授。主要研究方向为机器学习、数据挖掘、教育大数据分析、个性化推荐系统

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329420565888217339, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, xref=2, ext=[AuthorCompanyExt(id=1329420565930160380, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, companyId=1329420565888217339, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2School of Computer Science and Technology, University of Science and Technology of China, Hefei 230027), AuthorCompanyExt(id=1329420565938548989, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329420562385973463, companyId=1329420565888217339, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2中国科学技术大学计算机科学与技术学院 合肥 230027)])])]
龙佳琪1 , 张所娟1, * , 郝文宁1, 陈泽君1, 孙新杰2, 3, 刘逸凡1, 陈恩红2
作者信息
  • 1中国人民解放军陆军工程大学指挥控制工程学院 南京 210007
  • 2中国科学技术大学计算机科学与技术学院 合肥 230027
  • 3六盘水师范学院计算机科学学院 贵州六盘水 553004
通讯作者:
作者简介:

作者贡献声明

龙佳琪负责模型设计并撰写论文;张所娟制定论文整体研究框架;郝文宁提出关键改进意见;陈泽君、孙新杰参与了部分实验方法的分析与讨论;刘逸凡协助完成了数据收集与初步整理工作;陈恩红统筹研究进度与团队协作,确定论文选题方向。

2002年生。硕士研究生。主要研究方向为智能教育、知识追踪

1983年生。博士,教授。主要研究方向为教育数据挖掘、知识追踪、个性化学习推荐

1971年生。博士,教授。主要研究方向为大规模高维数据压缩、作战效能评估

1998年生。硕士研究生。主要研究方向为智能教育、学习推荐

1987年生。博士,教授。主要研究方向为智能教育、自适应学习

2000年生。硕士研究生。主要研究方向为智能教育、数据挖掘

1968年生。博士,教授。主要研究方向为机器学习、数据挖掘、教育大数据分析、个性化推荐系统

CAKT-Agent: A Context-Adaptive Knowledge Tracing Agent Based on MoE
Jiaqi Long1 , Suojuan Zhang1, * , Wenning Hao1, Zejun Chen1, Xinjie Sun2, 3, Yifan Liu1, Enhong Chen2
Affiliations
  • 1School of Command and Control Engineering, Army Engineering University of PLA, Nanjing 210007
  • 2School of Computer Science and Technology, University of Science and Technology of China, Hefei 230027
  • 3School of Computer Science, Liupanshui Normal University, Liupanshui, Guizhou 553004
  • born in 2002. Master candidate. Her main research interests include intelligent education and knowledge tracing

    born in 1983. PhD, professor. Her main research interests include educational data mining, knowledge tracing, and personalized exercise recommendation

    born in 1971. PhD, professor. His main research interests include large-scale high-dimensional data compression and evaluation of operational effectiveness

    born in 1998. Master candidate. Her main research interests include intelligent education and exercise recommendation

    born in 1987. PhD, professor. His main research interests include intelligent education and adaptive learning

    born in 2000. Master candidate. His main research interests include intelligent education and data mining

    born in 1968. PhD, professor. His main research interests include machine learning, data mining, education big data analysis, and personalized recommendation system

出版时间: 2026-09-01 doi: 10.7544/issn1000-1239.202660421
文章导航
收藏切换

知识追踪作为智能辅导系统的核心任务,旨在通过分析学生历史行为数据预测其未来表现,为个性化学习路径规划提供依据。当前主流深度学习方法受限于知识边界,依赖单一特征拟合,普遍存在学习过程建模过度简化、情境适配能力不足、预测过程黑箱化等问题,输出结果缺乏可解释性,难以转化为可落地的教学参考。为此,提出了基于混合专家(mixture of experts,MoE)模型的情境自适应知识追踪智能体CAKT-Agent,构建统一记忆模块、学习智能体和伴学智能体三级协同架构:统一记忆模块实现学习数据、教育先验知识和中间状态的全局存储与共享;学习智能体基于教育测量学和认知科学定义5维情境,设计维度感知的混合专家架构,通过两级门控机制实现情境自适应融合;伴学智能体依托专家激活权重与教育理论规则,生成具备教育可理解性的学习诊断报告。真实数据集实验结果表明,所提模型在学生表现预测精度上优于主流模型,且具备良好的情境自适应能力与结果可理解性。

知识追踪  /  混合专家模型  /  智能体推理  /  情境自适应  /  个性化学习  /  多智能体协同

Knowledge tracing, as the core task of intelligent tutoring systems, aims to predict students’ future learning performance by analyzing their historical behavioral data, and provides a fundamental basis for personalized learning path planning. Current mainstream deep learning methods are constrained by knowledge boundaries and rely on single-feature modeling, often exhibiting issues such as overly simplified learning process modeling, inadequate contextual adaptability, and black-box prediction mechanisms that yield results lacking interpretability and difficult to translate into actionable teaching references. To address these challenges, we propose CAKT-Agent, a context-adaptive knowledge tracing agent based on mixture-of-experts model, and construct a three-tier collaborative architecture consisting of a unified memory module, a learning agent, and a companion agent: the unified memory module realizes global storage and sharing of learning data, educational priors and intermediate model states; The learning agent defines a five-dimensional context based on educational measurement theory and cognitive science, designs a hybrid expert architecture with dimension-aware capabilities, and achieves context-adaptive integration through a two-level gating mechanism; the companion agent generates learning diagnostic reports with educational comprehensibility based on expert activation weights and educational theory rules. Experimental results on real-world datasets show that the proposed model outperforms mainstream knowledge tracing models in the prediction accuracy of students' learning performance, with excellent context-adaptive capability and result comprehensibility.

knowledge tracing  /  hybrid expert model  /  agent reasoning  /  contextual adaptation  /  personalized learning  /  multi-agent collaboration
龙佳琪, 张所娟, 郝文宁, 陈泽君, 孙新杰, 刘逸凡, 陈恩红. CAKT-Agent:基于MoE的情境自适应知识追踪智能体. 计算机研究与发展, 2026 , 63 (9) : 2319 -2337 . DOI: 10.7544/issn1000-1239.202660421
Jiaqi Long, Suojuan Zhang, Wenning Hao, Zejun Chen, Xinjie Sun, Yifan Liu, Enhong Chen. CAKT-Agent: A Context-Adaptive Knowledge Tracing Agent Based on MoE[J]. Journal of Computer Research and Development, 2026 , 63 (9) : 2319 -2337 . DOI: 10.7544/issn1000-1239.202660421
随着智能辅导系统(intelligent tutoring system,ITS)的普及,计算机技术与传统教育领域的深度结合已成为驱动教育智能化的核心引擎,吸引了众多研究人员关注。知识追踪(knowledge tracing,KT)作为动态感知学生认知状态、实现精准个性化指导的关键技术,逐渐成为推动智能教育发展的重要研究方向[1]。
知识追踪通过分析学习者的历史学习行为数据(如习题作答记录、学习时长、操作行为等),动态建模其知识掌握水平与认知状态演变,进而预测其在未来习题上的答题表现[2],为学习路径规划、资源推荐提供决策支撑[3]。早期知识追踪研究以基于隐马尔可夫模型(hidden Markov model,HMM)的贝叶斯知识追踪(Bayesian knowledge tracing,BKT)[4]模型为典型代表,为知识追踪提供了可解释的理论基础。随着深度学习技术的发展,基于神经网络的知识追踪不断涌现:Piech等人[5]提出的深度知识追踪(deep knowledge tracing,DKT)引入循环神经网络(recurrent neural network,RNN),开启了深度学习在该领域的应用;后续基于图神经网络(graph neural network,GNN)的GKT[6]、基于自注意力机制的SAKT[7]、融合项目反应理论(item response theory, IRT)的注意力知识追踪(AKT)[8]以及学习过程一致性模型LPKT[9]等相继被提出,从不同角度优化了模型性能,进一步推动知识追踪技术在智能辅导系统、在线教育平台等场景的落地应用。然而,随着学习场景的复杂化与学习者需求的多样化,如今的知识追踪方法面临着诸多新的挑战[10]。
图1展示了基于情境感知的知识追踪示例。智能辅导系统所收集的学生练习行为日志,涵盖带有知识概念标签的习题、学生作答对错结果、专家标注的习题难度及答题时长等多类异构信息,模型需从这些异构数据中提取有效情境特征,构建认知状态预测模型,进而预测在后续问题上的答题结果。事实上,学生在新问题上的作答情况往往高度依赖其历史学习行为。为了全面捕捉学习行为,需结合历史问题对应的知识概念、掌握情况及相关情境因素进行预测。
然而,传统知识追踪模型普遍对真实学习过程进行简化建模,多围绕固定的单一特征,将复杂学习过程建模为单一维度的知识状态更新[1]。简化学习者的交互行为,定义为$ X=\{({q}_{1},{r}_{1}),({q}_{2},{r}_{2}),…,({q}_{T},{r}_{T})\} $,仅聚焦答题结果与问题相关知识概念的二元关联,忽视了情境信息的多维性与动态性;部分情境感知知识追踪模型虽尝试融入部分情境特征,例如DIMKT[11]聚焦习题难度建模,LPKT[9]侧重时间维度特征挖掘,AKT[8]基于心理测量学IRT构建习题与知识点嵌入,MLC-DKT[12]设计习题-知识概念多层情境表征。但缺乏自适应的特征融合与建模机制,难以应对不同学习场景、不同学生个体的情境差异,导致模型在真实复杂学习场景中的泛化能力与预测精度受限。
此外,知识追踪任务的核心目标在于服务教学实践,教育工作者与学生需要的并非数值预测,而是具备逻辑的可理解结论[13]。当前知识追踪模型通常输出数值得分预测,无法直接为教学实施干预、个性化指导提供依据[14]。近年生成式智能体技术为该问题提供了新思路。如2T-Agent[15]构建类教师思维的认知诊断框架,通过两级反思机制生成因果解释链条;TRAVER[16]将知识追踪与逐轮验证相结合,先通过知识追踪估计学生知识状态,再通过验证机制引导辅导智能体完成任务。但现有方法缺乏对教育规则的显式记忆与可调用机制,解释过程依赖大语言模型的隐式参数记忆,无法将教育测量学、认知心理学等教育先验知识作为可验证的推理依据,导致诊断结果缺乏理论支撑。
基于此,本文提出基于混合专家的情境自适应知识追踪智能体(context-adaptive knowledge tracing agent based on MoE,CAKT-Agent),实现面向真实学习场景的情境自适应知识追踪与可理解性教学诊断。首先,定义多维度学习情境,设计专门架构捕获多维情境特征嵌入;以维度感知的混合专家模型为核心情境自适应单元,为每类情境分别配置专属的前馈神经网络(feed-forward neural network,FFN)专家与注意力专家资源池,并设计维度内混合门控与跨维度稀疏门控两级自适应机制,实现情境特征的自适应融合与学生认知状态的动态建模;构建由统一记忆模块、学习智能体和伴学智能体组成的三级协同智能体架构,以混合专家(mixture of experts,MoE)模型的情境感知特征与专家激活权重作为智能体交互的核心纽带,实现情境自适应能力在数据存储、逻辑推理、评估反馈全流程的赋能。
总之,本文的贡献有3个方面:
1)定义多维度学习情境。明确习题难度、知识关联强度、答题时长、错误模式及学习增益5类核心情境,设计专门特征提取架构,突破单一情境特征建模的局限,实现对多维情境的全面挖掘与嵌入表征。
2)提出维度感知的MoE架构。为每类情境配置专属的FFN专家与注意力专家资源池,设计维度内混合门控与跨维度稀疏门控两级自适应机制,精准捕捉不同情境对学生认知状态的差异化影响,有效适配复杂真实学习场景。
3)构建三级协同智能体架构。构建统一记忆模块、学习智能体和伴学智能体的协同框架,突破现有模型的数值预测局限,围绕情境信息解析预测逻辑,结合教育理论规则生成可理解的自然语言分析。
知识追踪的核心是通过学习者历史交互数据建模认知状态并预测未来表现。传统知识追踪模型,如BKT[4]模型通过估计答题结果的二元隐状态更新模型。随着深度神经网络技术发展,基于深度学习的知识追踪模型凭借序列建模优势成为研究主流。基于循环神经网络的DKT[5]模型是引入深度学习的模型,基于RNN对学习者的认知状态进行建模。动态键值记忆网络模型(DKVMN)[17]使用外部记忆矩阵来表示学习者的认知状态,进一步拓展了DKT模型的能力。基于图神经网络的GKT[6]、AKT模型[8]进一步拓展知识追踪的技术更新。CIKT[18]、SINKT[19]、LLM-KT[20]利用大语言模型的推理能力进一步拓展知识追踪的边界。
近年来,研究者开始关注情境感知的知识追踪模型,聚焦将学习行为置于真实动态环境中分析。教育学理论强调知识不能脱离具体情境而孤立存在,知识是在具体、可感知的情境中得以存在和理解的,只有通过在具体情境中的实践活动,知识才能被真正理解和接纳[21-22]。情境被定义为影响学习效果的环境因素,这些因素包括与教学内容相关联的、能够影响学习者绩效的各种信息[23]。早期情境建模的探索多为隐性的,AKT模型[8]将单调性注意力机制与心理测量学中Rasch模型结合,并利用题目和概念的嵌入正则化提升个性化表示能力。CAKQN模型[24]进一步扩展AKT,提出上下文感知的知识查询网络,将神经网络与心理测量理论结合,动态捕捉学习者与题目的交互情境。后续,研究者开始尝试定义情境。ContextKT[25]模型明确定义了5种情境(当前知识状态、记忆遗忘、问题-知识相关性、行为相关性、问题相关性),并整合多源上下文数据,显著提升了预测精度。MLC-DKT[12]模型通过分层结构捕捉学习过程中的多层次上下文信息,如行为模式、时间间隔等,增强对知识状态变化的刻画。研究表明,融入情境信息的知识追踪模型在性能和可解释性方面都具有一定的优势[10]。
然而,以上研究主要集中在习题和知识点相关的情境信息,且部分情境不具备普适性[26]。因此,本文调研现有知识追踪模型,选取了已知建模探索比较充分的情境作为本文情境自适应的探索重点。
MoE模型由Jacobs等人[27]于1991年提出,核心架构包含多个专业化的专家网络与一个全局门控网络,二者协同实现“分而治之”的任务适配。早期MoE模型主要应用于传统神经网络的过拟合缓解与小型任务的性能优化[28]。近年来,出现了Mixtral of Experts[29]、Mixture of A Million Experts[30]等新型变体,进一步推动了其在多领域的普及。
随着知识追踪任务对异质性特征建模需求的提升,MoE因其多专家协同、自适应分工的核心优势,逐渐被应用于知识追踪领域。Liao等人提出了RouterKT[31],引入以人为本的路由机制,将多头注意力机制作为专家。该研究验证了MoE在捕捉学习者个体差异、提升预测精度上的有效性,然而这种尝试为通用序列MoE的直接迁移,未考虑教育情境异构性的语义边界,出现专家任务混杂、特征耦合问题;单一门控机制无法兼顾单情境局部时序关联与多情境全局竞争关系,导致模型泛化能力不足。
近年来,基于大语言模型(large language model,LLM)的智能体为教育领域模拟现实教育场景提供了新机遇[32-33]。LLM的推理能力应用于教育领域的多个方面。如Li等人[34]提出的认知引导智能体,通过LLM解析学习者答题序列,生成结构化的知识状态报告,不仅能输出掌握程度评分,还能解释薄弱环节形成原因;CIKT[18]通过LLM构建知识点关联图谱,强化知识迁移效应建模;思维链提示的LLM智能体[35],可根据学习者知识状态动态调整讲解逻辑,实现“诊断—推荐—辅导”的闭环服务;此外Agent4Edu[36]代表了LLM智能体模拟学生的最新框架。然而其设计针对通用场景的学生模拟,不适合对具体学生的认知状态进行模拟和探究。本文通过关注学生历史行为对情境信息的敏感程度,复制认知过程行动策略,并给予针对性的解答。此外,为应对智能体设计缺乏清晰的功能分工以及易出现模块边界模糊、功能重叠的问题,本文设计了追踪学习者认知状态的学习智能体和生成可理解学习诊断报告的伴学智能体。
在线学习系统中,知识追踪的核心任务是通过学习者历史作答序列建模其认知状态,进而预测未来作答表现。定义$ U=\{{u}_{1},{u}_{2},… ,{u}_{I}\} $为包含$ I $个学习者的学生集。学习者$ {u}_{i} $在$ T $时间步内作答序列为$ {X}^{i}=\{(q_{1}^{i},r_{1}^{i},\mathcal{C}_{1}^{i}),(q_{2}^{i},r_{2}^{i},\mathcal{C}_{2}^{i}),… ,(q_{T}^{i},r_{T}^{i},\mathcal{C}_{T}^{i})\} $,其中$ q_{t}^{i} $为学习者$ {u}_{i} $在时间步$ t $回答的问题,对应特定的知识概念集合。$ r_{t}^{i}\in \{0,1\} $代表学习者$ {u}_{i} $回答问题$ q_{t}^{i} $的结果,1为正确作答,0为错误作答。$ \mathcal{C}_{t}^{i} $为学习者$ {u}_{i} $在时间步$ t $作答时的相关情境信息,是补充刻画作答场景与个体学习特质的核心信息。
通过学习者历史作答序列$ {X}^{i} $(含历史问题、作答结果、情境信息的三元组)和新问题$ q_{T+1}^{i} $,知识追踪任务可定义为条件概率 $ p\left(r_{T+1}^{i}\mid q_{T+1}^{i},{X}^{i}\right) $,即基于历史问题、作答结果及情境信息与新问题,预测新问题作答结果为$ r_{T+1}^{i} $的概率。
学习情境是指与学习行为相关,在学生学习行为交互过程中体现的上下文信息,是情境认知理论中影响学习者知识构建与认知发展的核心环境变量[21, 37]。在具体定义时调研现有知识追踪模型,结合教育场景的实际需求与现有研究基础,选取已知建模探索充分的情境作为本文情境自适应的研究重点。具体如下:
1)习题难度($ {f}_{1} $)。基于教育测量理论,习题难度指习题通过自身解题复杂度和知识概念深度对学习者产生的个性化影响程度[38-39]。难度适配可强化学习效果,难度过高易致认知过载,难度过低则难以巩固知识。
2)知识关联强度($ {f}_{2} $)。在实际学习场景中,知识点并非孤立存在,存在前置、后继或相似等关联,学习者对某一知识点的掌握情况会直接影响其对关联知识点的作答表现[6,40]。
3)答题时长($ {f}_{3} $)。答题时长能够间接反映知识掌握熟练度,即掌握良好者答题时长较短且稳定;对未掌握或模糊者往往需要更长时间思考,甚至出现答题时长异常波动的情况[9]。
4)错误模式($ {f}_{4} $)。错误模式由学习者在不同时间步对各知识点的答题正确率表示,可揭示捕捉其知识掌握的变化趋势与薄弱环节[41]。
5)学习增益($ {f}_{5} $)。学习增益是衡量学习者通过练习、复习获得的知识提升[42]。结合艾宾浩斯遗忘曲线,学习者对知识点的记忆会随时间衰减,通过练习、复习等交互行为可实现记忆强化与知识增益[43]。
上述5类情境分别为题目属性、知识关系、行为时间、表现序列、学习进步5个维度,覆盖了教育心理学经典的“输入—过程—输出”核心环节,也契合已有知识追踪研究中对多维上下文信息的建模实践(情境信息相关计算公式见附录A)。
CAKT-Agent由统一记忆模块、学习智能体和伴学智能体三大部分协同构成,以维度感知的MoE模型为核心,实现统一记忆存储、精准认知建模与个性化反馈。统一记忆模块作为全局数据中心,负责存储过程数据、教育规则及反思结果。学习智能体作为底层建模支撑,负责执行认知状态追踪与预测,持续监测预测偏差。伴学智能体作为用户交互入口,接收自然语言查询,生成结构化的诊断结论。
在运行流程上,系统首先加载学生历史学习数据,调用学习智能体中的MoE情境自适应工具与LoRA微调LLM预测器执行前向传播,生成情境感知特征与专家激活权重并缓存至统一记忆模块。当接收到用户查询时,伴学智能体依次完成问题理解、情境推理和响应生成3个阶段,最终返回自然语言响应。
如图2所示,CAKT-Agent框架由3个主要部分组成:统一记忆模块、学习智能体和伴学智能体。
统一记忆模块是CAKT-Agent得以实现数据共享与知识复用的基础,避免每次推理请求触发耗时的模型全量重计算。模块严格遵循认知心理学领域经典的长时记忆三元分类通用框架,该框架由Tulving[44]提出,将人类长时记忆划分为情境记忆、语义记忆与程序性记忆3类,是教育学、认知科学领域被广泛认可并沿用至今的人类学习机制核心理论[45]。基于此,本文将存储内容划分为情境记忆、语义记忆和程序性记忆,为智能体的感知推理提供了全方位的记忆支撑。
1)情境记忆。用于存储个体亲历的、带有时间戳与场景信息的事件性经历[46],在本研究中存储学习者原始行为轨迹与模型建模过程数据,是智能体完成认知建模的基础。对于任意学习者$ {u}_{i} $,情境记忆不仅记录交互记录原始字段,并实时同步存储学习智能体前向传播产生的高维中间变量。具体而言,在时间步$ t $,学习智能体完成对当前交互序列的特征提取与预测后,MoE情境自适应工具生成的情境感知特征$ {\boldsymbol{H}}_{t} $、专家激活权重$ {\boldsymbol{\alpha }}_{t}=({\alpha }_{t1},{\alpha }_{t2},{\alpha }_{t3},{\alpha }_{t4},{\alpha }_{t5}) $以及微调LLM预测器输出的预测概率$ {P}_{\text{final}} $被序列化并作为元数据记录至学生的情境记忆条目。形式化表示为:
$ {\mathcal{M}}_{\text{episodic}}({u}_{i},t)=\left\{{\boldsymbol{H}}_{t},{\boldsymbol{\alpha }}_{t},{P}_{\text{final}},{q}_{t},{r}_{t},{C}_{t}\right\} , $
其中$ {q}_{t} $、$ {r}_{t} $、$ {C}_{t} $分别为习题、答题结果和情境信息,$ {\boldsymbol{\alpha }}_{t} $的各分量对应5维情境信息相对重要性度量,其数值反映不同情境维度在本次认知状态更新中的贡献度,由此构成了伴学智能体情境归因与可理解推理的核心数值依据。此外,为平衡存储开销与推理实时性,情境记忆使用基于滑动窗口的缓存淘汰机制,保留最近100条MoE输出记录。
2)语义记忆。用于存储与个体无关的、通用的事实性知识、概念体系与逻辑规则[44],本文中存储的与具体学生个体无关的通用性教育领域先验知识与逻辑规则,是保证伴学智能体生成诊断结论具备科学性与教育可理解性的理论基石。其结构化内容可形式化为二元组$ {\mathcal{M}}_{\text{semantic}}=(\mathcal{F},\mathcal{R}) $,其中$ \mathcal{F} $为MoE专家语义映射函数,$ \mathcal{R} $为可执行诊断规则库。
① MoE专家语义映射表$ {\mathcal{F}}_{}\colon \{1,2,3,4,5\}\rightarrow \mathcal{D} $是将学习智能体中抽象的数字专家索引翻译为教育工作者和学生能够直观理解的描述性术语,$ \mathcal{D} $为语义标签集合。 例如专家1被定义为“习题难度专家”,其激活程度表征当前习题的客观难度;专家2被定义为 “知识关联强度专家”,其激活程度表征前置知识点掌握情况对当前作答的影响程度,以此类推完成5类专家的全量语义映射。
② 可执行的诊断规则库。由$ N $条基于经典教育测量理论(项目反应理论 IRT)、认知负荷理论、艾宾浩斯遗忘曲线理论凝练而成的标准化推理规则构成,每条规则定义为从专家权值与答题结果到归因结论的映射:$ \mathcal{R}=\left\{{R}_{k}\colon {\varPhi }_{k}(\boldsymbol{\alpha },r)\mapsto {{Con}}_{k}\right\}_{k=1}^{N} $,其中$ \boldsymbol{\alpha }\in {\mathbb{R}}^{5} $为专家激活权重向量,$ r\in \{0,1\} $为作答结果。$ {\varPhi }_{k}(\boldsymbol{\alpha },r) $为规则$ k $的触发条件谓词,通常表现为专家权重的阈值判定及其与答题结果的逻辑组合。伴学智能体的情境推理引擎通过遍历$ \mathcal{R} $,匹配满足$ {\varPhi }_{k}(\boldsymbol{\alpha },r)=\text{True} $的规则,将数值权重模式转化为具有明确教育含义的可解释诊断陈述。
3)程序性记忆。在经典认知理论中,核心用于存储“如何完成任务”的过程性、操作性知识,是个体技能习得、认知能力迭代进化的核心载体,ACT-R认知理论进一步将其作为学习者从陈述性知识向程序性技能转化的核心存储单元[47]。在本研究中程序性记忆为CAKT-Agent实现自我反思与持续进化的元认知存储单元,记录学习智能体中反思器产生的所有偏差分析记录、参数优化历史以及长期统计规律。与情境记忆关注个体交互细节和语义记忆关注静态通用知识不同,程序性记忆捕捉的是模型在认知建模过程中习得的迭代优化规律,是实现模型持续进化、诊断结论置信度校准的核心枢纽。具体来说,包含3个子类:
① 偏差事件日志。保存每次预测结果与真实作答出现显著偏差的完整事件记录,包含偏差发生的时间点、对应习题与知识点、偏差幅度、当前专家激活权重及反思器分析得出的核心成因。
② 优化轨迹。记录反思器向MoE门控网络、LLM质量感知门控每次发出的微调指令及其效果验证,形成可追溯的演化路径。
③ 长期偏置统计表。汇总模型在特定知识点上表现出的系统性倾向,包含知识点标识、系统性偏置类型(高估/低估)、平均偏差幅度及偏差发生次数。
程序性记忆可形式化为:
$ {\mathcal{M}}_{\text{procedural}}=({\mathcal{L}}_{\text{dev}},{\mathcal{T}}_{\text{opt}},{\mathcal{S}}_{\text{bias}}) \text{,} $
其中,$ {\mathcal{L}}_{\text{dev}} $为偏差事件日志,$ {\mathcal{T}}_{\text{opt}} $为优化轨迹,$ {\mathcal{S}}_{\text{bias}} $为长期偏置统计表。
程序性记忆数据在伴学智能体生成最终诊断建议时被作为置信度修正因子加以利用。当检测到对某知识点的评估存在历史性高估偏置时,系统会在报告中附加提示:“模型对该部分掌握情况的评估可能偏于乐观,建议结合近期课堂观察综合判断”。
学习智能体作为CAKT-Agent的建模核心,负责处理学生历史交互序列,以输出高精度的预测,同时提供全流程交互的先验数据支撑。将学生零散、非结构化的历史交互数据,通过MoE情境自适应工具转化为可量化、可理解、能反映学生真实认知状态的情境感知特征,并使用微调LLM预测器预测结果,同时反思器实现模型的自我纠错、自我优化,确保认知建模的准确性与时效性,为推理与反馈提供可靠的数据支撑。具体来说,学习智能体由3个核心模块组成:MoE情境自适应工具、微调LLM预测器及反思器。
MoE情境自适应工具是学习智能体的核心建模单元,负责执行MoE前向传播并提取情境感知特征与专家激活权重。习题难度、知识关联强度、答题时长、错误模式以及学习增益这5维情境因素对学生认知状态的影响机制并不相同,若简单拼接后建模会导致特征间相互干扰与重要信息损失。本文设计了维度感知的MoE架构,首先依据附录A中公式,为每个时间步分别提取5个独立的情境特征,随后分别输入至5个参数独立的专属专家网络$ {E}_{1} $~$ {E}_{5} $中深度加工,获取高层次抽象表征$ {\boldsymbol{h}}_{i}={E}_{i}({f}_{i})\left(i=1,… ,5\right) $。
对于第$ i $个情境维度,其内部采用混合门控机制在FFN专家与注意力专家之间动态选择。以序列全局均值$ {\overline{f}}_{i}=\displaystyle\sum\limits_{t=1}^{T}f_{i}^{(t)}/{T} $与当前状态$ {f}_{i} $的拼接作为输入,混合门控输出二维分支权重:
$ {\boldsymbol{\beta }}_{i}={Softmax}\left(\boldsymbol{W}_{i}^{{{\mathrm{mix}}}}\cdot [{\overline{f}}_{\mathrm{i}}\oplus {f}_{i}]+{\boldsymbol{b}}_{i}^{{{\mathrm{mix}}}}\right)=(\beta _{i}^{{FFN}},\beta _{{i}}^{{Attn}}) \text{,} $
其中,FFN分支内配置$ {K}_{f} $个专家,通过内部门控计算权重$ \boldsymbol{\gamma }_{i}^{{FFN}}={Softmax}(\boldsymbol{W}_{i}^{f}{f}_{i}+\boldsymbol{b}_{i}^{f}) $,分支输出为$ \boldsymbol{h}_{i}^{{FFN}}= \displaystyle\sum\limits_{k=1}^{{K}_{FFN}}\gamma _{i,k}^{{{{FFN}}}}\cdot {{FFN}}_{k}({f}_{i}) $;注意力分支同理,输出$ \boldsymbol{h}_{i}^{{Attn}}= \displaystyle\sum\limits_{k=1}^{{K}_{Attn}}\gamma _{i,k}^{{Attn}}\cdot {{Attn}}_{k}({f}_{i}) $,注意力计算采用因果掩码。最终,经加权混合与残差连接后层归一化得到:
$ {\boldsymbol{h}}_{i}={LayerNorm}\left({f}_{i}+\beta _{i}^{{FFN}}\cdot \boldsymbol{h}_{i}^{{FFN}}+\beta _{i}^{{Attn}}\cdot \boldsymbol{h}_{i}^{{Attn}}\right) \text{,} $
5维情境表征拼接后送入跨维度门控网络,计算专家激活权重$ \boldsymbol{\alpha }\in {\mathbb{R}}^{5} $:
$ \boldsymbol{\alpha }={Softmax}\left({\boldsymbol{W}}_{\alpha}\cdot [{\boldsymbol{h}}_{1}\oplus {\boldsymbol{h}}_{2}\oplus {\boldsymbol{h}}_{3}\oplus {\boldsymbol{h}}_{4}\oplus {\boldsymbol{h}}_{5}]+{\boldsymbol{b}}_{\alpha}\right) \;。 $
为提升聚焦能力,模型支持跨维度Top-K稀疏选择。令$ \boldsymbol{\alpha } $中最大的$ {K} $个分量保留,其余置零后重新归一化得到稀疏权值$ \widetilde{\boldsymbol{\alpha }} $。最终MoE融合输出为:
$ \boldsymbol{H}=\sum \limits_{i=1}^{5}{\tilde{\alpha }}_{i}\cdot {\boldsymbol{h}}_{i}\;。$
模型采用分而治之与动态融合结合的机制,根据不同学习场景自适应调整各特征权重。专家激活权重$ \boldsymbol{\alpha } $与融合后情境感知特征$ \boldsymbol{H} $作为元数据写入情境记忆,为伴学智能体提供观察模型内部决策依据的窗口。
为进一步强化对学习者长期行为依赖与复杂交互模式的捕捉,引入微调LLM预测器作为增强模块。采用LoRA轻量化适配策略,冻结LLM的原始参数,仅在注意力层$ Q $、$ \mathcal{K} $、$ \mathcal{V} $、$ \mathcal{O} $投影旁添加低秩分解矩阵$ \boldsymbol{B}\in {\mathbb{R}}^{d\times r} $和$ \boldsymbol{A}\in {\mathbb{R}}^{r\times d}\left(r=8\right) $,使可训练参数量大幅降低。前向传播时,MoE的输出$ \boldsymbol{H} $先经时序建模层$ \boldsymbol{Z}={MultiHeadAttn}(\boldsymbol{H}) $保障底线预测能力,随后压缩投影映射至LLM词嵌入维度$ \widetilde{\boldsymbol{Z}}={\boldsymbol{W}}_{\widetilde{\boldsymbol{Z}}}\boldsymbol{Z} $。$ \widetilde{\boldsymbol{Z}} $输入至冻结的LLM,提取最后3层隐状态并通过可学习权重$ {\boldsymbol{w}}_{l}\in {\mathbb{R}}^{3} $进行加权融合:
$ {\boldsymbol{H}}_{\text{LLM}}=\sum \limits_{l=1}^{3}{{\boldsymbol{w}}}_{l}\cdot {{LLM}}_{l}(\widetilde{\boldsymbol{Z}}) \;。 $
质量感知门控根据LLM输出方差动态调节融合权重,以抑制LLM不确定性带来的噪声。计算LLM表征方差$ \sigma _{\text{LLM}}^{2}={Var}({\boldsymbol{H}}_{\text{LLM}}) $与时序建模表征方差 $ \sigma _{\text{kt}}^{2}= {Var}(\boldsymbol{Z}) $,门控系数定义为$ g=\exp (-\sigma _{\text{LLM}}^{2}/\tau ) $,其中$ \tau $为温度参数。最终融合表征与预测概率为:
$ {\boldsymbol{H}}_{\text{final}}=g\cdot {\boldsymbol{H}}_{\text{LLM}}+(1-g)\cdot \boldsymbol{Z}\text{,}{P}_{\text{final}}=\sigma (\boldsymbol{w}_{{\mathrm{o}}}^{{\mathrm{T}}}{\boldsymbol{H}}_{\text{final}}) \text{,} $
其中,$ {\boldsymbol{H}}_{\text{final}} $为经质量感知门控的情境感知特征,$ {P}_{\text{final}} $为最终预测结果,$ \sigma \left(\cdot \right) $为Sigmoid激活函数,$ \boldsymbol{w}_{{\mathrm{o}}}^{{\mathrm{T}}} $为权值矩阵。
若LLM加载失败或推理异常,系统自动回退至备用预测头$ {P}_{\text{fallback}}=\sigma ({MLP}(\boldsymbol{Z})) $以确保服务可用。
模型训练的总损失函数由主任务损失与2项辅助损失加权构成。主任务损失为二元交叉熵损失$ {\mathcal{L}}_{\text{main}} $,用来衡量模型预测概率与真实答题结果之间的差异。训练阶段引入2项辅助损失:跨维度负载均衡损失$ {\mathcal{L}}_{\text{bal}} $鼓励权值均匀分布,维度内混合均衡损失$ {\mathcal{L}}_{\text{mix}} $促使专家类型均衡使用,与主任务损失加权相加。公式如下:
$ {\mathcal{L}}_{\text{main}}=-\left[{r}_{t}\log P_{\text{final}}^{(t)}+(1-{r}_{t})\log (1-P_{\text{final}}^{(t)})\right] \text{,} $
$ {\mathcal{L}}_{\text{bal}}={Var}(\boldsymbol{\alpha })=\frac{1}{5}\sum \limits_{i=1}^{5}{\left({\alpha }_{i}-\frac{1}{5}\right)}^{2} \text{,} $
$ {{\cal L}_{{\rm{mix}}}} = - \sum\limits_{i = 1}^5 {\sum\limits_{m \in \{ {{FFN}},{{Attn}}\} } {\beta _i^m} } \log \beta _i^m\;, $
$ {\mathcal{L}}_{\text{total}}={\mathcal{L}}_{\text{main}}+{\lambda }_{1}{\mathcal{L}}_{\text{bal}}+{\lambda }_{2}{\mathcal{L}}_{\text{mix}} \text{,} $
其中$ {\lambda }_{1} $和$ {\lambda }_{2} $为辅助损失的权重系数,参考领域已有研究[48],结合辅助损失约束特性设置$ {\lambda }_{1}={\lambda }_{2}=0.01 $。总损失在训练阶段驱动MoE情境自适应工具、微调LLM预测器以及门控网络的端到端优化,确保模型在追求预测精度的同时能均衡利用多维情境信息。
一个具备自我纠错与持续进化能力的系统才具备在复杂真实环境中长期部署的价值,为此学习智能体中特别设计了反思器。反思器扮演着模型数据监控者的角色,不直接参与前向预测,而是分析预测结果与真实反馈之间的偏差反向指导模型参数的优化方向。反思器内部维护着一个偏差检测触发器,其核心逻辑为计算预测概率$ {P}_{{\mathrm{final}}} $与学生真实答题结果$ r $之间的绝对差值,当该差值超过一个经过实验验证的最优阈值时,偏差分析流程被激活,即$ | $$ {P}_{\text{final}}-r| \gt \delta $。一旦触发,反思器不会全局更新参数,而是分析偏差发生时刻产生的专家激活权重$ \boldsymbol{\alpha } $。通过计算各专家权重与偏差幅度之间的相关性,反思器能够将预测失误精准归因于1个或多个情境维度的建模不足。例如,反思器发现如果一系列错误预测均伴随难度专家权重$ {\alpha }_{{\mathrm{diff}}} $的异常偏低,则生成一条定向优化指令,指示MoE门控网络在处理后续类似难度题目时,提高对难度特征的敏感度。同理,若偏差被归因于对长时遗忘效应的捕捉失败,反思器将建议调整微调LLM预测器的权重,从而更依赖LLM的长序列记忆能力。
除触发式的参数微调指令外,反思器还会以批次为单位对偏差模式进行宏观统计,分别计算模型对每位学生的高估频次与低估频次,以及模型在每个知识点上的系统性预测偏置。这些统计数据被周期性地写入统一记忆模块的程序性记忆中,为后续伴学智能体在生成诊断报告时评估结论的可靠性提供重要参考。
伴学智能体作为CAKT-Agent与外部用户(包括教师与学生)进行信息交换的交互中枢,成为连接后端精准认知建模能力与前端用户个性化需求的桥梁,其核心设计目标是接收用户以自然语言方式提出的开放性问题,通过标准化的语义解析、动态推理调度与个性化表达生成,为师生提供智能伴学服务。伴学智能体的内部工作流程由3个串行且相互衔接的处理阶段构成,分别是任务解析与规划、情境推理引擎以及响应生成与风格学习,旨在实现对用户从模糊意图表达到具体文本输出的全链路支持。
作为交互起始点,模块在接收到用户输入自然语言查询$ Q $后,首先基于正则表达式识别与关键词匹配以较低的计算开销判断查询是否属于系统支持的教育场景范畴,并初步识别其任务大类,过滤与学习诊断无关的闲聊或无效输入。而后,调用LLM进行深度语义理解与实体抽取,生成一份标准化的任务分析报告。该报告的数据结构包括:任务类型、关键实体列表,以及为解决该问题所需执行的子任务序列。
$ \mathcal{T}={TaskAnalyzer}(Q;\mathcal{P},{\theta }_{\text{LLM}}) \;。 $
该阶段会生成包含多个子任务的探索规划,引导后续的情境推理引擎分步执行数据检索与逻辑推导,从而避免因单次推理信息不足而导致的回答片面或肤浅。
情境推理引擎阶段是实现从数值权重向教学洞见质变转化的关键环节。根据任务指令中指定的目标学生与时间范围,向统一记忆模块发起精确的数据检索请求。从情境记忆中拉取目标学生在相关时间点上的专家权重向量序列以及对应的答题结果序列。同时,引擎从语义记忆中加载相关的教育先验知识。
以归因分析为例,答错时专家权重向量为(0.1, 0.8, 0.05, 0.03, 0.02),引擎发现$ {\alpha }_{\text{sso}}=0.8> 0.7 $,命中规则“IF $ {\alpha }_{{\mathrm{sso}}}> 0.7 $ THEN知识关联强度为主导情境因素”。引擎进一步演绎出该学生失败源于无法有效调动前置知识模块来输出结构化诊断结论。最终,情境推理引擎阶段依据子任务生成的结构化推理结论生成结构化学习报告,并调用程序记忆修正结论置信度。若存在偏差,查找偏差原因并给出相应建议。
在获得学习报告后,由自然语言生成与风格学习阶段负责将这份报告渲染为自然语言文本。该模块的设计目标不仅仅是完成信息的无损转述,更追求与用户建立一种自然、舒适且高效的交流体验。系统会分析用户的历史提问与反馈,从正式度、详尽度、词汇难度3个维度持续更新该用户的风格画像[49-50]。如一位习惯于使用专业术语的教师,其正式度得分会趋近于1;而一位喜欢简短直接反馈的学生,其详尽度得分则会较低。当需要生成回复时,会根据用户风格画像构建高度定制化的提示词,不仅包含了情境推理引擎输出的结构化学习报告,还嵌入了对语言风格的显式约束指令,例如“请用鼓励、亲切的语气”或“请保持客观、严谨的学术口吻”。随后,该提示词被发送至LLM,生成过程被严格控制以保持回复的简洁性,单次响应长度通常不超过150字,从而避免信息过载,确保沟通的高效性。
在实验过程中,借鉴pyKT[51]的处理方式对学习记录进行系统化处理。输入序列长度统一为200,零向量填充,排除记录少于30条的样本。选取数据集的20%作为测试集,数据集均遵循标准5倍交叉验证流程进行预处理。为全面验证模型性能,实验覆盖单步预测与多步预测两大核心任务,其中多步预测同时实现非累积式与累积式2种主流预测策略;评估过程严格遵循pyKT的防标签泄露规范,采用all-in-one题目级评估范式,并选取pyKT模型库中8种主流经典与前沿知识追踪模型作为对比基线,确保实验结果的公平性、可复现性与横向可比性。
模型训练阶段,学习率初始为0.002,每5个训练周期衰减50%以保障模型收敛,特征嵌入维度参数统一设置为128。实验基于Linux操作系统,硬件环境为16 vCPU AMD EPYC 9K84 96-Core Processor、150 GB内存与H20-NVLink显卡,模型骨干采用经LoRA轻量化微调的Llama-3.2-8B大语言模型1,其中LoRA秩rank=8,缩放参数ɑ=32,训练采用早停策略,验证集AUC连续10个训练周期无提升时终止训练。
便于对比,将MoE情境自适应工具、微调LLM预测器集成为CAKT,用于验证知识追踪预测性能。
为了全面验证本文提出模型的有效性,选用了2个具有代表性的真实在线教育数据集进行实验,剔除了存在缺失数据的记录,统计结果如表1所示。
1)NIPS34数据集2源自2020年NeurIPS教育挑战赛中的任务3和任务4,包含学生对多项选择题形式的数学诊断题答案,数据采集自Eedi平台。
2)ASSIST2012数据集3是ASSISTments在线智能数学辅导系统在2012—2013学年收集的,是该系列规模最大、应用最广泛的数据集之一。
1)DKT[5]是最早基于深度学习的知识追踪模型,利用RNN/LSTM网络评估学生认知状态掌握程度。
2)DKVMN[17]是基于存储网络的知识追踪模型,定义了存储学生知识状态的矩阵值。
3)AKT[8]是一种具有上下文感知能力的注意力知识追踪模型。该模型通过注意力机制构建知识检索器,实时捕捉学生动态知识状态。
4)RobustKT[52]通过平滑和减法操作将认知模式与随机因素分离,采用门控或融合技术对它们进行重新加权,在噪声环境下增强知识追踪效果。
5)LPKT[9]通过直接模拟真实学生学习过程,依托自适应时序门控捕捉知识累积增益与自然衰减,以学习过程一致性约束提升状态追踪真实性。
6)DIMKT[11]将难度级别融入问题表示中,并在练习过程中建模学习者认知状态与习题难度级别关系。探索了问题难度效应,展现了卓越的可解释性。
7)LefoKT[53]通过相对遗忘注意力机制,将遗忘模式与问题相关性解耦,并优化时序门控的衰减权重,更准确地模拟交互序列中的多样化遗忘行为。
8)UKT[54]将学生知识状态建模为连续概率分布,采用Wasserstein注意力衡量分布间距离,引入不确定性感知对比学习量化答题置信度,以不确定性视角优化知识状态估计。
1)AUC(area under curve)。AUC是ROC曲线下面积,用于衡量模型排序能力。
$ {{AUC}} = \dfrac{{\displaystyle\sum\limits_{i \in P} {\displaystyle\sum\limits_{j \in N} } \;{\mathbb{I}}(scor{e_i} \gt scor{e_j})}}{{|P| \cdot |N|}}, $
其中,$P $为正样本集合,$N$为负样本集合。${\mathbb{I}}(\cdot) $为指示函数,条件成立取 1,否则 0。${\mathbb{I}}(\cdot) $取值为[0,1],越接近1效果越好。
2)准确率(accuracy, ACC)。ACC用于计算样本预测正确的占比,全部样本里预测和真实标签一致的比例。
$ {{ACC}} = \frac{{TP + TN}}{{TP + TN + FP + FN}}, $
其中,$TP $为真正例,$TN $为真负例,$FP $为假正例,$FN $为假负例。ACC取值为[0,1],越接近1效果越好。
3)均方根误差(root mean square error, RMSE)。RMSE是回归任务指标,预测值和真实值之间误差的平方根。用于预测学生掌握度、预测分数等连续输出。
$ {{RMSE}} = \sqrt {\frac{1}{n}\sum\limits_{i = 1}^n {{{({y_i} - {{\hat y}_i})}^2}} }, $
其中,$y_i $为真实标签(连续值),${{{\hat y}_i}} $为模型预测值,$n $为样本总数。$RESE\geqslant 0 $,数值越小代表预测越准。
表2和表3分别展示了各模型在单步预测与非累计式多步预测知识追踪任务上的性能对比。在NIPS34数据集上,单步预测实验中模型最优实验CAKT的AUC较次优基线AKT提升8.83%,ACC提升8.39%,RMSE降低10.86%;非累积式多步预测中,CAKT的AUC较现有最优模型AKT仍提升1.45%,ACC提升1.37%,RMSE降低0.99%。在ASSIST2012数据集上,CAKT的各指标性能也为最优。这一卓越表现充分表明,多维度学习情境的全面捕捉、维度感知的MoE设计是必要且极具价值的。
为进一步展示累积式多步预测的动态建模能力,选取测试集中70%的真实学习交互作为观测历史,模型从第一个未观测习题开始逐题预测其答对概率,并将每一步预测结果离散化后反馈至后续输入。该过程模拟了真实教学场景中系统根据学生连续表现动态更新知识状态的过程,实验结果如图3所示。其平均逐时间步准确率达到72.34%,高于所有对比模型;相较最强基线DIMKT,CAKT仍取得约0.88%的提升。进一步分阶段分析发现,随着预测步长不断增加,DKT、UKT、LPKT等基线模型的预测误差持续大幅上升,存在严重的长时序性能漂移问题;CAKT模型整体波动幅度远小于所有对比方法,即使在长时序阶段准确率也超基线模型。这一现象说明,CAKT在个体学生长周期连续学情追踪场景下具备更优异的预测稳定性与鲁棒性,更适配真实在线教学系统动态推演学生未来答题表现的业务场景。
本实验验证CAKT模型各核心组件有效性。其中,消融组件对应命名规则如下:A1~A5分别表示移除习题难度、移除知识关联强度、移除答题时长、移除错误模式、移除学习增益;B1表示移除LLM,B2表示移除LoRA;C表示移除维度感知;D表示移除门控。实验结果如图4所示,CAKT模型的各模块对预测性能均有贡献,且模块重要性因数据集特点存在差异。
在NIPS34数据集上LLM模块影响最显著,移除后AUC下降0.0555、ACC下降5.12个百分点、RMSE上升0.0381,因该数据集知识密度高,LLM的语义理解能力可有效挖掘习题与知识的深层关联;维度感知的MoE模块次之,消融后指标明显退化,验证其对5维特征的精准融合作用。在ASSIST2012数据集上,知识关联强度和错误模式模块影响更突出,消融后AUC分别下降0.0141、0.0133,因该数据集样本异质性强,二者可提升模型对不同答题场景的适配性。5维特征中,答题时长对模型性能影响最小,而学习增益、习题难度可辅助提升预测精度。综上,CAKT模型的各模块协同作用,可有效适配不同特点的数据集,其中情境感知的MoE模型和LLM是保障模型高性能的核心模块。
为验证CAKT模型核心创新模块的稳定性与参数敏感性,本实验围绕模型关键参数TOP-K(专家选择数量)和专家池数量开展敏感性测试,保持其余超参数与消融实验一致,以AUC、ACC为评价标准,分析参数变化对模型性能的影响,结果如图5所示。
在TOP-K参数测试中,NIPS34数据集性能随TOP-K增大逐步提升,在TOP-5时达到最优,而TOP-1时性能最差(AUC = 0.8331),原因在于该数据集知识关联紧密,需更多专家参与特征融合才能挖掘深层关联。ASSIST2012数据集则在TOP-3时达到性能最优(AUC = 0.7987、ACC = 0.7682),TOP-K中K>3后性能略有下降,因数据集样本异质性强,过多专家参与易引入冗余信息。在专家池数量测试中,2个数据集均在专家池数量为20时达到最优性能,专家池数量为10时因专家资源不足导致性能下降,专家池数量为40时因冗余专家增加计算成本且降低融合精度,说明模型创新的维度感知机制可在最优专家池规模下实现高效特征融合。
为验证不同LLM及其LoRA微调版本在 CAKT框架中的适配性能,本文对比了BERT、Llama-3.2-1B、Llama-3.2-3B和Llama-3.2-8B四种模型在无LLM、基线LLM、LoRA微调LLM三个阶段的性能表现,结果如图6所示。
整体来看,LLM融合可显著提升知识追踪任务性能。与无LLM的基础CAKT模型相比,所有基线LLM在3项指标上均实现了一致且大幅的优化。其中 Llama-3.2-8B 基线模型提升幅度达5.55%,验证了LLM能够有效挖掘学生答题序列中隐含的知识状态演化规律。Llama系列模型整体优于BERT模型,且性能随参数量增加呈小幅上升趋势,然而三者性能差距不足0.2个百分点,说明在知识追踪任务中,LLM参数量的边际收益已趋于平缓,中等规模模型即可达到接近最优的性能。此外,LoRA微调对不同模型的提升效果呈现显著差异化特征:Llama-3.2-3B、Llama-3.2-8B微调后性能有小幅提升,Llama-3.2-1B微调后性能基本无变化,推测是由于小模型容量有限,已达到该数据集的性能上限,进一步微调无法带来额外增益。
为评估CAKT-Agent在真实在线场景中的部署可行性,实验分别统计了离线训练阶段与在线服务阶段的资源消耗。离线训练阶段衡量模型构建成本(训练耗时、参数量、Checkpoint大小、峰值显存);在线服务阶段面向单条学生学情诊断请求(输入序列长度不大于200),统计平均推理延迟、推理峰值显存、端到端总耗时与Token消耗,完整对比数据如表4所示。
离线训练阶段属于上线前一次性前置投入,不代表线上并发推理的运行负载。传统轻量化知识追踪基线(DKT、AKT等)结构简单,迭代耗时与显存占用均较低。在NIPS34数据集上,本文使用轻量化CAKT-LLM,不启用完整LLM主干和LoRA,AUC较次优模型仍提升2.11%,效率优势明显。在加载全量CAKT-Agent模型时,峰值显存达36113.31 MB,这主要源于多专家并行分发与长序列LLM特征融合。但该部分开销仅存在于上线前一次性训练流程。同时,通过LoRA微调,CAKT-Agent的可训练参数仅4.08×106,Checkpoint主要来自基础LLM的冻结权重,实际可部署时仅需加载4.08×106的增量权重和少量推理引擎,存储与加载成本大幅降低。
离线训练为一次性投入,在线推理时延、显存占用才是在线教育平台并发服务的核心评判指标。基于统一标准学生交互样本开展线上推理实测:轻量化 CAKT-LLM单请求平均推理延迟仅50.44 ms,推理峰值显存47.92 MB,推理响应稳定维持在毫秒级,完全满足在线教育系统实时交互的业务时延要求;即便搭载完整LLM基座的CAKT-Agent,单请求平均推理延迟也仅90.44 ms,线上推理算力开销增幅可控。综合可见,轻量化CAKT-LLM方案可直接适配中小算力线上服务集群,完整CAKT-Agent亦可在常规GPU环境下稳定并发运行。
为系统验证 CAKT-Agent 生成的诊断报告的可信度与泛化能力,本文设计并实施了3项实验:1)真实案例分析;2)诊断报告质量评估;3)人工诊断一致性分析。从案例分析到教育专家评分诊断,验证结果可理解性。
案例分析以NIPS34数据集的学生494为研究对象,基于CAKT-Agent模型全流程推理能力,演示伴学智能体的实现流程。实验以学生494的问题为输入,展示了从加载学生数据、前向传播,到伴学智能体处理问题、给出回答的全流程,如图7所示。
首先,任务解析与规划阶段分析外部用户提问,解析为任务分析报告,划分了3个子任务。而后,在情境引擎分析阶段,调取前向传播记录的权重分配,精准定位学生核心情境,成功识别出学生494的核心影响因素为习题难度(权重为0.38)和错误模式(权重为0.21)。结合专家分支诊断,进一步锁定新知识点难度梯度过陡、核心知识点存在理解障碍的核心问题。融合LLM增强推理与规则引擎的创新架构,在错误原因分析中,通过19条规则匹配与LLM深度推理,精准挖掘出错误根源(新知识点难度无铺垫、前置知识不扎实)。基于上述结论,LLM总结生成覆盖子任务的学习报告。最后,响应生成与风格学习阶段结合学习报告,根据用户输入调整输出风格,提升了模型的实际应用适配性。
为深入验证本文提出的CAKT-Agent诊断结果的可理解性和准确性,实验邀请3位在计算机教育和教育领域拥有具有5年以上研究或教学经验的教育专家,对20名抽样学生的4类诊断报告进行独立盲审。4类报告分别为:CAKT-Agent;w/o Memory(移除统一记忆模块);w/o Reflection(移除反思器);Direct LLM(直接调用GLM-4-flash)。每位专家从诊断准确性、可理解性、教学可实现性3个维度对每份报告进行1~5分量表评分(评分表详情在附录B)。实验结果如图8所示。
结果显示,CAKT-Agent在可理解性的均值最高,为4.55,综合均值为4.23。配对样本t检验显示,CAKT-Agent显著优于各消融变体:诊断准确性比Direct LLM高1.10分,可理解性高0.75分,2个对比的显著性均为p < 0.001。此外,w/o Reflection的教学可实现性得分最高(4.25),Direct LLM得分为3.95。CAKT-Agent的诊断准确性高于w/o Reflection,说明反思器的偏差复核机制能够提升诊断准确性。移除统一记忆模块后,诊断准确性和综合均值均下降,表明该模块存储的教育领域先验知识与专家激活权重是生成个性化诊断的关键,缺少该模块会导致诊断针对性不足。移除反思器后,诊断准确性下降更明显,但教学可实现性略有提升,这说明反思器的偏差复核机制的主要作用是校验诊断准确性,抑制诊断偏差。
为补充主观评分的不足,本文进一步度量了系统诊断标签与专家独立标注之间的一致性。3位专家分别对20名学生独立标注薄弱知识点、错误模式、学习状态和教学干预策略4类结构化标签,与 CAKT-Agent输出的对应字段计算一致性指标。结果如表5所示。
CAKT-Agent在所有诊断维度上均与专家标注取得了较高一致性。薄弱知识点识别的精度Precision为0.807、召回率Recall为0.761,表明系统能够有效定位学生的知识薄弱环节。错误模式的F1最高达到 0.927,说明系统对“概念未掌握”“遗忘”“粗心波动”“迁移困难”等错误归因与专家判断高度吻合。
教学干预策略的Precision为0.885,但Recall相对较低(0.600),F1为0.715。Recall偏低的主要原因在于专家标注的干预策略覆盖面更广(如额外标注“分层提升练习”“学习节奏调整”等),而系统输出的策略未能完全覆盖,导致部分专家标签未被匹配。
学习状态分类的ACC为65%,不一致的案例主要集中在“基础薄弱型”与“退步型”及“波动型”与“高掌握稳定型”之间的混淆,反映出系统在学习趋势判断上与专家存在一定分歧。总体而言,CAKT-Agent的诊断输出与人类专家判断具有较高一致性。系统可作为教师决策的有效辅助参考,但在学习状态的精细分类上仍需结合教师经验进行综合判断。
本文提出了 CAKT-Agent,一种基于MoE的情境自适应知识追踪智能体。该模型包含统一记忆模块、学习智能体和伴学智能体,以维度感知的MoE模型与两级自适应门控机制为核心创新点,结合LoRA轻量化微调的LLM增强推理,在提升预测性能的同时,具备良好的教育可理解性。
具体而言,本文明确定义了习题难度、知识关联强度、答题时长、错误模式及学习增益5类核心学习情境,设计专属特征提取架构,突破单一情境建模的局限;提出维度感知的MoE架构,为每类情境配置专属FFN专家与注意力专家资源池,通过维度内混合门控与跨维度稀疏门控两级机制,实现情境特征的自适应融合与学生认知状态的动态建模;构建统一记忆模块并遵循长时记忆三元分类框架,实现数据全局共享与知识复用;设计学习智能体完成认知状态追踪与自我优化,伴学智能体实现自然语言交互与个性化诊断反馈,打破现有模型的数值预测局限。
为完整验证CAKT-Agent整体性能,本文基于NIPS34、ASSIST2012两大公开真实在线教育数据集开展多维度实验,包含知识追踪预测性能对比和教育专家诊断评测等,从预测精度、推理效率、诊断可信度多角度验证模型优势。
未来研究可围绕以下三大方向开展:
一是拓展多维学习情境输入,当前仅聚焦习题与学生自身交互特征,后续可融入教师指导、课堂环境、同伴互动等外部环境变量,进一步提升复杂线下、线上混合学习场景下的情境自适应建模能力。
二是扩充多场景、多学科验证数据集,完成跨领域泛化性验证,推进CAKT-Agent规模化落地至商用智能辅导系统。
三是研究分层轻量化部署方案,适配不同算力等级在线教育平台与边缘终端设备:①MoE专家稀疏裁剪优化,固定最优TOP-K路由数量,剔除冗余情境专家降低前向计算开销;②轻量化大模型基座适配,采用Llama-3.2-1B/3B小模型替代8B基座,搭配LoRA极小参数量微调平衡精度与算力;③统一记忆冷热分层缓存,高频活跃学生数据内存常驻、低频历史数据磁盘持久存储,降低在线并发场景显存占用,为中小机构、移动端轻量化智能辅导应用提供可行落地路径。

参考文献 引证文献
排序方式:
1
Abdelrahman G, Wang Qing, Nunes B. Knowledge tracing: A survey[J]. ACM Computing Surveys, 2023, 55(11): 1−37
2
刘铁园, 陈威, 常亮, 等. 基于深度学习的知识追踪研究进展[J]. 计算机研究与发展, 2022, 59(1): 81−104
Liu Tieyuan, Chen Wei, Chang Liang, et al. Research advances in the knowledge tracing based on deep learning[J]. Journal of Computer Research and Development, 2022, 59(1): 81−104 (in Chinese)
3
Shen Shuanghong, Liu Qi, Huang Zhenya, et al. A survey of knowledge tracing: Models, variants, and applications[J]. IEEE Transactions on Learning Technologies, 2024, 17: 1858−1879
4
Corbett A T, Anderson J R. Knowledge tracing: Modeling the acquisition of procedural knowledge[J]. User Modeling and User-Adapted Interaction, 1994, 4(4): 253−278
5
Piech C, Bassen J, Huang J, et al. Deep knowledge tracing[J]. Advances in Neural Information Processing Systems, 2015, 28:1−12
6
Nakagawa H, Iwasawa Y, Matsuo Y. Graph-based knowledge tracing: Modeling student proficiency using graph neural network[C]//Proc of IEEE/WIC/ACM Int Conf on Web Intelligence. New York: ACM, 2019: 156−163
7
Pandey S, Karypis G. A self-attentive model for knowledge tracing[J]. arXiv preprint, arXiv: 1907.06837, 2019
8
Ghosh A, Heffernan N, Lan A S. Context-aware attentive knowledge tracing[C]//Proc of the 26th ACM SIGKDD Int Conf on Knowledge Discovery &amp; Data Mining. New York: ACM, 2020: 2330−2339
9
Shen Shuanghong, Liu Qi, Chen Enhong, et al. Learning process-consistent knowledge tracing[C]//Proc of the 27th ACM SIGKDD Conf on Knowledge Discovery &amp; Data Mining. New York: ACM, 2021: 1452−1460
10
Zhu Siyi, Chen Wenjie. Deep learning based knowledge tracing: A review of the literature[C]//Proc of the 2025 Int Conf on Big Data and Informatization Education. New York: ACM, 2025: 81−87
11
Shen Shuanghong, Huang Zhenya, Liu Qi, et al. Assessing student’s dynamic knowledge state by exploring the question difficulty effect[C]//Proc of the 45th Int ACM SIGIR Conf on Research and Development in Information Retrieval. New York: ACM, 2022: 427−437
12
Zhang Suojuan, Pu Jie, Cui Jing, et al. MLC-DKT: A multi-layer context-aware deep knowledge tracing model[J]. Knowledge-Based Systems, 2024, 303: 112384
13
Bai Yanhong, Zhao Jiabao, Wei Tingjiang, et al. A survey of explainable knowledge tracing[J]. arXiv preprint, arXiv: 2403.07279, 2024
14
Voultsiou E, Moussiades L. A systematic review of AI, VR, and LLM applications in special education: Opportunities, challenges, and future directions[J]. Education and Information Technologies, 2025, 30(13): 19141−19181
15
孙新杰, 刘淇, 张凯, 等. 2T-Agent: 生成式智能体驱动的类教师思维认知诊断框架[J]. 计算机学报, 2026, 49(3): 574−590
Sun Xinjie, Liu Qi, Zhang Kai, et al. 2T-Agent: A generative agent-driven cognitive diagnostic framework for teacher-like thinking[J]. Chinese Journal of Computers, 2026, 49(3): 574−590 (in Chinese)
16
Wang Jian, Dai Yinpei, Zhang Yichi, et al. Training turn-by-turn verifiers for dialogue tutoring agents: The curious case of LLMs as your coding tutors[C]//Findings of the Association for Computational Linguistics: ACL 2025. Stroudsburg, PA: ACL, 2025: 12416−12436
17
Zhang Jiani, Shi Xingjian, King I, et al. Dynamic key-value memory networks for knowledge tracing[C]//Proc of the 26th Int Conf on World Wide Web. New York: ACM, 2017: 765−774
18
Li Runze, Wu Siyu, Wang Jun, et al. CIKT: A collaborative and iterative knowledge tracing framework with large language models[J]. arXiv preprint, arXiv: 2505.17705, 2025
19
Fu Lingyue, Guan Hao, Du Kounianhua, et al. Sinkt: A structure-aware inductive knowledge tracing model with large language model[C]//Proc of the 33rd ACM Int Conf on Information and Knowledge Management. New York: ACM, 2024: 632−642
20
Wang Ziwei, Zhou Jie, Chen Qin, et al. LLM-KT: Aligning large language models with knowledge tracing using a plug-and-play instruction[J]. arXiv preprint, arXiv: 2502.02945, 2025
21
Brown J S, Collins A, Duguid P. Situated cognition and the culture of learning[J]. Educational Researcher, 1989, 18(1): 32−42
22
Lave J, Wenger E. Situated Learning: Legitimate Peripheral Participation[M]. Cambridge, UK: Cambridge University Press, 1991
23
蒲杰, 张所娟, 陈卫卫. 知识情境感知的深度知识追踪模型[J]. 南京大学学报 (自然科学版), 2024, 60(1): 76−86
Pu Jie, Zhang Suojuan, Chen Weiwei. A deep knowledge tracking model for contextual awareness of knowledge[J]. Journal of Nanjing University (Natural Sciences Edition), 2024, 60(1): 76−86 (in Chinese)
24
Cheng Yan, Wu Gang. Zou Haifeng, et al. A knowledge query network model based on rasch model embedding for personalized online learning[J]. Frontiers in Psychology, 2022, 13: 846621
25
Yu Minghe, Li Fan, Liu Hengyu, et al. ContextKT: A context-based method for knowledge tracing[J]. Applied Sciences, 2022, 12(17): 8822
26
Yamkovenko B, Hogg C A R, Miller-Vedam M, et al. Practical evaluation of deep knowledge tracing models for use in learning platforms[J/OL]. International Educational Data Mining Society, 2025[2026-05-29]. https://educationaldatamining.org/EDM2025/Proc/2025.EDM.industry-papers.46/
27
Jacobs R A, Jordan M I, Nowlan S J, et al. Adaptive mixtures of local experts[J]. Neural Computation, 1991, 3(1): 79−87
28
Fedus W, Zoph B, Shazeer N. Switch Transformers: Scaling to trillion parameter models with simple and efficient sparsity[J]. Journal of Machine Learning Research, 2022, 23(120): 1−39
29
Jiang A Q, Sablayrolles A, Roux A, et al. Mixtral of experts[J]. arXiv preprint, arXiv: 2401.04088, 2024
30
He X O. Mixture of a million experts[J]. arXiv preprint, arXiv: 2407.04153, 2024
31
Liao Han, Zu Shuaishuai. RouterKT: Mixture-of-experts for knowledge tracing[J]. arXiv preprint, arXiv: 2504.08989, 2025
32
Nikolova E. Artificial intelligence tools into higher mathematics education: Opportunities, challenges, and student perceptions[J]. Mathematics and Informatics, 2024, 67(4): 406−417
33
陈慧敏, 刘知远, 孙茂松. 大语言模型时代的社会机遇与挑战[J]. 计算机研究与发展, 2024, 61(5): 1094−1103
Chen Huimin, Liu Zhiyuan, Sun Maosong. The social opportunities and challenges in the era of large language models[J]. Journal of Computer Research and Development, 2024, 61(5): 1094−1103 (in Chinese)
34
Li Haoxuan, Yu Jifan, Ouyang Yuanxin, et al. Explainable few-shot knowledge tracing[J]. Frontiers of Digital Education, 2025, 2(4): 34
35
Berthon A, van der Schaar M. Language bottleneck models: A framework for interpretable knowledge tracing and beyond[J]. arXiv preprint, arXiv: 2506.16982, 2025
36
Gao Weibo, Liu Qi, Yue Linan, et al. Agent4edu: Generating learner response data by generative agents for intelligent education systems[C]//Proc of the AAAI Conf on Artificial Intelligence. Palo Alto, CA: AAAI, 2025, 39(22): 23923−23932
37
高文. 情境学习与情境认知[J]. 教育发展研究, 2001(8): 30−35
Gao Wen. Situational learning and situational cognition[J]. Educational Development Research, 2001(8): 30−35 (in Chinese)
38
Lord F M. Applications of Item Response Theory to Practical Testing Problems[M]. Abingdon-on-Thames, UK: Routledge, 2012
39
Sweller J. Cognitive load theory, learning difficulty, and instructional design[J]. Learning and Instruction, 1994, 4(4): 295−312
40
Ausubel D P, Novak J D, Hanesian H. Educational Psychology: A Cognitive View[M]. New York: Holt, Rinehart and Winston, 1978
41
Ghosh A, Raspat J, Lan A. Option tracing: Beyond correctness analysis in knowledge tracing[C]//Proc of Int Conf on Artificial Intelligence in Education. Cham, Switzerland: Springer, 2021: 137−149
42
Karpicke J D, Roediger III H L. The critical importance of retrieval for learning[J]. Science, 2008, 319(5865): 966−968
43
Nagatani K, Zhang Qian, Sato M, et al. Augmenting knowledge tracing by considering forgetting behavior[C]//Proc of the World Wide Web Conf. New York: ACM, 2019: 3101−3107
44
Tulving E. Episodic and semantic memory[J]. Organization of Memory, 1972, 1(1): 381−403
45
Park J S, O'Brien J C, Cai C J, et al. Generative agents: Interactive simulacra of human behavior[C]//Proc of the 36th Annual ACM Symp on User Interface Software and Technology. New York: ACM, 2023: 1−22
46
隋洁, 吴艳红. 心理时间之旅——情景记忆的独特性[J]. 北京大学学报(自然科学版), 2004(2): 326−332
Sui Jie, Wu Yanhong. A psychological journey through time: The uniqueness of situational memory[J]. Journal of Peking University (Natural Sciences Edition), 2004(2): 326−332 (in Chinese)
47
Anderson J R. Acquisition of cognitive skill[J]. Psychological Review, 1982, 89(4): 369−406
48
许嘉, 唐嵘蓉, 吕品, 等. 基于学习迁移的稳定知识追踪模型[J]. 华南师范大学学报(自然科学版), 2024, 56(4): 68−79
Xu Jia, Tang Rongrong, Lü Pin, et al. A stable knowledge tracking model based on learning transfer[J]. Journal of South China Normal University (Natural Sciences Edition), 2024, 56(4): 68−79 (in Chinese)
49
Wang Qingsong, Yao Chang, Chen Jingyuan. Causal-Steer: Disentangled continuous style control without parallel corpora[C/OL]//Proc of the 14th Int Conf on Learning Representations. [2026-05-29]. https://openreview.net/pdf?id=yfiOYXvsX5
50
Alhafni B, Kulkarni V, Kumar D, et al. Personalized text generation with fine-grained linguistic control[C]//Proc of the 1st Workshop on Personalization of Generative AI Systems (PERSONALIZE 2024). Stroudsburg, PA: ACL, 2024: 88−101
51
Liu Zitao, Liu Qiongqiong, Chen Jiahao, et al. pyKT: A Python library to benchmark deep learning based knowledge tracing models[J]. Advances in Neural Information Processing Systems, 2022, 35: 18542−18555
52
Guo Teng, Qin Yu, Xia Yubin, et al. Enhancing knowledge tracing through decoupling cognitive pattern from error-prone data[C]//Proc of the ACM on Web Conf 2025. New York: ACM, 2025: 5108−5116
53
Bai Youheng, Li Xueyi, Liu Zitao, et al. Rethinking and improving student learning and forgetting processes for attention based knowledge tracing models[C]//Proc of the AAAI Conf on Artificial Intelligence. Palo Alto, CA: AAAI, 2025, 39(27): 27822−27830
54
Cheng Weihua, Du Hanwen, Li Chunxiao, et al. Uncertainty-aware knowledge tracing[C]//Proc of the AAAI Conf on Artificial Intelligence. Palo Alto, CA: AAAI, 2025, 39(27): 27905−27913
2026年第63卷第9期
74
40
0
引用本文
BibTeX
文章信息
doi: 10.7544/issn1000-1239.202660421
  • 接收时间:2026-06-06
  • 首发时间:2026-11-17
  • 出版时间:2026-09-01
补充材料
相关文章
文章信息
作者
出版历史
  • 收稿日期:2026-06-06
  • 修回日期:2026-07-15
基金
作者信息
    1中国人民解放军陆军工程大学指挥控制工程学院 南京 210007
    2中国科学技术大学计算机科学与技术学院 合肥 230027
    3六盘水师范学院计算机科学学院 贵州六盘水 553004

通讯作者:

参考文献
分享链接
https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/10.7544/issn1000-1239.202660421
分享至
全文二维码

扫描看全文

引用本文
BibTeX
本文的引用情况
2种不同金属材料的力学参数
科
Family
属数
Number of
genus
种数
Number of
species
占总种数比例
Percentage of
total species (%)
属
Genus
种数
Number of
species
占总种数比例
Percentage of total
species (%)
鹅膏菌科Amanitaceae 2 11 5.26 鹅膏菌属 Amanita 10 4.78
小菇科 Mycenaceae 2 12 5.74 丝盖伞属 Inocybe 5 2.39
多孔菌科 Polyporaceae 8 14 6.70 蜡蘑属 Laccaria 5 2.39
红菇科 Russulaceae 3 23 11.00 小皮伞属 Marasmius 6 2.87
小菇属 Mycena 11 5.26
光柄菇属 Pluteus 5 2.39
红菇属 Russula 17 8.13
栓菌属 Trametes 5 2.39
关闭全屏