To address the problems of traditional early childhood development assessments, such as over-reliance on teachers' subjective observations, limited data sources, and difficulty in reflecting children's continuous growth, this paper proposes an interpretable multimodal large language model (LLM) method for early childhood development assessment based on multi-source structured data. A multimodal feature fusion mechanism is designed, and an early childhood development assessment model is established through cross-modal alignment, dynamic weight allocation, and time-series information integration. Furthermore, attention mechanisms, feature contribution analysis, and inference chain interpretation techniques are introduced to enhance the transparency and credibility of the model's assessment results. Combined with dynamic modeling of developmental trajectories and growth trend analysis, the model achieves process-oriented assessment and prediction of children's comprehensive development. Experimental results show that the proposed model achieves 95.1% accuracy and 94.1% F1-score in early childhood development assessment tasks, representing improvements of 2.8% and 2.8% compared to GPT-4o, and improvements of 4.9% and 5.2% compared to the multimodal Transformer, respectively. These results demonstrate that the proposed interpretable multimodal LLM assessment model can achieve accurate assessment, reliable interpretation, and dynamic tracking of children's comprehensive development, providing a new technical path for personalized education support and scientific decision-making in early childhood.

Xiaowen Wang, Wanyi Huang, Xiao Liu, Zhengfen Teng, Yaoyao Cai. Interpretable Multimodal LLM Evaluation Model for Comprehensive Development of Young Children Based on Multi-Source Structured Data. Frontiers in Educational Research (2026), Vol. 9, Issue 6: 97-106. https://doi.org/10.25236/FER.2026.090613.

[1] Destriana R, Aksani M L, Priyanggodo D Y, et al. Design of a Digital Platform for PAUD Child Development Monitoring Using the Dynamic Systems Development Method and Machine Learning[J]. Jurnal Teknik Informatika (Jutif), 2025, 6(5): 3587-3601.

[2] Chen W, Shen Z, Pan Y, et al. Applying machine learning algorithm to optimize personalized education recommendation system[J]. Journal of Theory and Practice of Engineering Science, 2024, 4(01): 101-108.

[3] Chen X, Hu Z, Wang C. Empowering education development through AIGC: A systematic literature review[J]. Education and Information Technologies, 2024, 29(13): 17485-17537.

[4] Ouyang F, Dinh T A, Xu W. A systematic review of AI-driven educational assessment in STEM education[J]. Journal for STEM Education Research, 2023, 6(3): 408-426.

[5] Luo Y, Han X, Zhang C. Prediction of learning outcomes with a machine learning algorithm based on online learning behavior data in blended courses[J]. Asia Pacific Education Review, 2024, 25(2): 267-285.

[6] Wang S, Xu T, Li H, et al. Large language models for education: A survey and outlook[J]. IEEE Signal Processing Magazine, 2026, 42(6): 51-63.