Abstract：Computerized adaptive testing(CAT) is designed to achieve efficient assessment through dynamic question selection. However, existing methods still suffer from insufficient accuracy in semantic modeling and ability estimation. To address these issues, a text-semantic enhanced graph neural network based CAT approach(TECAT) is proposed. Contextual semantic representations of questions and concepts are extracted using a pretrained language model, and multi-level dependencies among questions and concepts are captured by applying graph attention networks on the question-concept and concept-prerequisite graphs. To integrate semantic and structural information, a gated fusion mechanism based on additive attention and SiLU activation is introduced for the above two types of information to be adaptively combined. As a result, more expressive node representations are obtained. The CAT process is further formulated as a multi-objective reinforcement learning task to jointly optimize question quality, diversity, and novelty. A quality reward function based on changes in ability estimation error is designed to better reflect the contribution of each question to ability diagnosis. Experiments on two real-world datasets, Eedi and Junyi, show that TECAT achieves superior ability in estimation accuracy and concept representation quality compared with the existing methods.
[1] WAINER H, DORANS N J, FLAUGHER D, et al. Computerized Adaptive Testing: A Primer. 2nd Edition. New York, USA: Routledge, 2000. [2] LIU Q, ZHUANG Y, BI H Y, et al. Survey of Computerized Adaptive Testing: A Machine Learning[C/OL].[2025-07-11]. https://arxiv.org/pdf/2404.00712. [3] LIU Y J, ZHANG T C, WANG X C, et al. New Development of Cognitive Diagnosis Models. Frontiers of Computer Science, 2022, 17(1). DOI: 10.1007/s11704-022-1128-3. [4] LU H R, WANG L, MA X L, et al. A Survey of Graph Neural Networks and Their Industrial Applications. Neurocomputing, 2025, 614. DOI: 10.1016/j.neucom.2024.128761. [5] WANG H Y, LONG T, YIN L, et al. GMOCAT: A Graph-Enhanced Multi-objective Method for Computerized Adaptive Testing// Proc of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. New York, USA: ACM, 2023: 2279-2289. [6] VRAHATIS A G, LAZAROS K, KOTSIANTIS S.Graph Attention Networks: A Comprehensive Review of Methods and Applications. Future Internet, 2024, 16(9). DOI: 10.3390/fi16090318. [7] RISSANEN J J.Fisher Information and Stochastic Complexity. IEEE Transactions on Information Theory, 1996, 42(1): 40-47. [8] CHANG H H, YING Z L.A Global Information Approach to Computerized Adaptive Testing. Applied Psychological Measurement, 1996, 20(3): 213-229. [9] CAI L, CHOI K, HANSEN M, et al. Item Response Theory. Annual Review of Statistics and Its Application, 2016, 3(1): 297-321. [10] BI H Y, MA H P, HUANG Z Y, et al. Quality Meets Diversity: A Model-Agnostic Framework for Computerized Adaptive Testing// Proc of the IEEE International Conference on Data Mining. Washing-ton, USA: IEEE, 2020: 42-51. [11] GHOSH A, LAN A. BOBCAT: Bilevel Optimization-Based Computerized Adaptive Testing// Proc of the 30th International Joint Conference on Artificial Intelligence. San Francisco, USA: IJCAI,2410-2417. [12] HU K, LI M Y, SONG Z Q, et al. A Review of Research on Reinforcement Learning Algorithms for Multi-agents. Neurocomputing, 2024, 599. DOI: 10.1016/j.neucom.2024.128068. [13] ZHUANG Y, LIU Q, HUANG Z Y, et al. Fully Adaptive Framework: Neural Computerized Adaptive Testing for Online Education. Proceedings of the AAAI Conference on Artificial Intelligence, 2022, 36(4): 4734-4742. [14] MNIH V, KAVUKCUOGLU K, SILVER D, et al. Playing Atari with Deep Reinforcement Learning[C/OL]. [2025-07-11]. https://arxiv.org/pdf/1312.5602. [15] HAARNOJA T, TANG H R, ABBEEL P, et al. Reinforcement Learning with Deep Energy-Based Policies// Proc of the 34th International Conference on Machine Learning. San Diego, USA: JMLR, 2017: 1352-1361. [16] WANG P J, LIU H Y, XU M Q.An Adaptive Testing Item Selection Strategy via a Deep Reinforcement Learning Approach. Behavior Research Methods, 2024, 56(8): 8695-8714. [17] FELTEN F, TALBI E G, DANOY G.Multi-objective Reinforcement Learning Based on Decomposition: A Taxonomy and Framework. Journal of Artificial Intelligence Research, 2024, 79: 679-723. [18] YIN Y, LIU Q, HUANG Z Y, et al. QuesNet: A Unified Representation for Heterogeneous Test Questions// Proc of the 25th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. New York, USA: ACM, 2019: 1328-1336. [19] NING Y T, HUANG Z Y, LIN X, et al. Towards a Holistic Understanding of Mathematical Questions with Contrastive Pre-trai-ning. Proceedings of the AAAI Conference on Artificial Intelligence, 2023, 37(11): 13409-13418. [20] XIAO S T, LIU Z, ZHANG P T, et al. C-Pack: Packed Resources for General Chinese Embeddings// Proc of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval. New York, USA: ACM, 2024: 641-649. [21] WANG F, LIU Q, CHEN E H, et al. Neural Cognitive Diagnosis for Intelligent Education Systems. Proceedings of the AAAI Confe-rence on Artificial Intelligence, 2020, 34(4): 6153-6161. [22] HAN K C T. Components of the Item Selection Algorithm in Computerized Adaptive Testing. Journal of Educational Evaluation for Health Professions, 2018, 15. DOI: 10.3352/jeehp.2018.15.7. [23] SCHULMAN J, WOLSKI F, DHARIWAL P, et al. Proximal Policy Optimization Algorithms[C/OL].[2025-07-11]. https://arxiv.org/pdf/1707.06347. [24] WANG Z C, LAMB A, SAVELIEV E, et al. Instructions and Guide for Diagnostic Questions: The NeurIPS 2020 Education Challenge[C/OL]. [2025-07-11]. http://arxiv.org/pdf/2007.12061. [25] CHANG H S, HSU H J, CHEN K T. Modeling Exercise Relationships in E-Learning: A Unified Approach[C/OL]. [2025-07-11].https://www.educationaldatamining.org/EDM2015/proceedings/short532-535.pdf. [26] KINGMA D P, BA J L. Adam: A Method for Stochastic Optimiza-tion[C/OL]. [2025-07-11]. https://arxiv.org/pdf/1412.6980. [27] GAO W B, LIU Q, HUANG Z Y, et al. RCD: Relation Map Dri-ven Cognitive Diagnosis for Intelligent Education Systems// Proc of the 44th International ACM SIGIR Conference on Research and Development in Information Retrieval. New York, USA: ACM, 2021: 501-510. [28] HAN H. The Utility of Receiver Operating Characteristic Curve in Educational Assessment: Performance Prediction. Mathematics, 2022, 10(9). DOI: 10.3390/math10091493. [29] ZHUANG Y, LIU Q, ZHAO G H, et al. A Bounded Ability Estimation for Computerized Adaptive Testing// Proc of the 37th International Conference on Neural Information Processing Systems. Cambridge, USA: MIT Press, 2023: 2381-2402. [30] YU J H, ZHUANG Y, HUANG Z Y, et al. A Unified Adaptive Testing System Enabled by Hierarchical Structure Search// Proc of the 41st International Conference on Machine Learning. San Diego, USA: JMLR, 2024: 57803-57817. [31] MCINNES L, HEALY J, MELVILLE J. UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction[C/OL]. [2025-07-11]. https://arxiv.org/pdf/1802.03426.
2025, Vol. 38 
