A Survey on AI-Powered Adaptive Learning Using Concept Mapping and Grounded Response Generation Technology

Abstract

The proliferation of digital educational resources has intensified the demand for intelligent systems capable of understanding, organizing, and presenting knowledge in ways that align with individual learner needs. Conventional e-learning environments continue to rely on passive content delivery mechanisms, including static document repositories, manualrevision scheduling,andkeyword-basedretrieval,allof which failtocapturethesemanticstructureofa subject domain or the evolving retention state of a learner. This survey systematically examines the theoretical and technical foundations underlying intelligent adaptive learning, spanning Concept Graph construction, retrieval-supported generation, dense vector search, neural knowledge tracing, spaced-repetition scheduling, and AI-driven tutoring. Building on this review, the paper introduces Neuroweave, a comprehensive adaptive learning framework that ingests uploaded educational documents, extracts domain concepts, and organizes them into a living Concept Graph. Memory retention is quantified per concept through theNeuroweave Adaptive MemoryAlgorithm (NAMA), which weighs review frequency, quiz performance, graph-level reinforcement from connected concepts, and time-based exponential decay. Grounded question answering is delivered by coupling large language model generation with vector retrieval over thelearner\'s own studymaterial, ensuring factual accuracy and source transparency. The paper presents the architectural design, component interactions, algorithmic formulation, comparative analysis against related literature, identified limitations, and directions for future research.

Introduction

This study presents Neuroweave, an AI-powered adaptive learning platform that combines Concept Graphs, Retrieval-Supported Generation (RSG), and memory modeling to create a personalized and intelligent learning environment. Traditional learning platforms are often static, providing fixed content sequences and generic feedback, which limits learning effectiveness. Neuroweave addresses these limitations by dynamically adapting learning paths based on each student's knowledge level, concept mastery, and memory retention.

The system uses Concept Graphs to represent relationships between academic concepts, enabling identification of prerequisite knowledge gaps and personalized study recommendations. RSG enhances the reliability of AI-generated answers by retrieving information directly from uploaded study materials, reducing hallucinations and ensuring factual accuracy. Additionally, the Neuroweave Adaptive Memory Algorithm (NAMA) models memory decay and retention, helping schedule reviews at optimal times to strengthen long-term learning.

Neuroweave features a modern architecture with a React-based frontend, FastAPI backend, PostgreSQL with pgvector, BGE embeddings for semantic retrieval, and the Llama-3.3-70B language model for concept extraction and question answering. The platform supports document ingestion, concept extraction, graph construction, adaptive study planning, memory heatmaps, and grounded question answering through its "Ask Your Brain" interface.

The literature review highlights advances in graph-based adaptive learning, retrieval-augmented generation, knowledge tracing, and semantic retrieval, while identifying challenges such as graph construction complexity, computational overhead, and privacy concerns. Compared to existing learning management systems, flashcard tools, and general AI assistants, Neuroweave provides deeper concept modeling, retention-aware scheduling, and factually grounded responses.

Conclusion

This survey has examined the theoretical landscape of AI-powered adaptive learning by reviewing foundational contributions in Concept Graph construction, retrieval-supportedgeneration,vectorsearch,neuralknowledgetracing, and memory modeling. Existing work demonstrates the individual value of each component but rarelyintegrates them into a cohesive, deployment-ready system accessible to individual learners without institutional infrastructure. Neuroweave addresses this gap by providing an end-to-end platform that transforms uploaded documents into a living ConceptGraph,computesper-conceptretentionscoresthrough NAMA, and grounds AI-generated answers in verified source material via vector retrieval and large language model generation. The system reduces learner effort for material organisation, increases revision targeting precision, and provides transparent, source-anchored responses to subject-matter queries. Avenues for future research include longitudinal validation of NAMA with diverse real-world learner cohorts, automated weightcalibrationthroughreinforcementlearningfromlearner outcome feedback, multimodal Concept Graph construction from lecture videos and diagrammatic content, privacy-preservingfederatedlearningacrossinstitutionaldeployments, andintegrationofreal-timecollaborativelearningfeaturesthat allow peer Concept Graphs to inform individual study recommendations.

References

[1] A.Lietal.,Electronics,vol.15,no. 1,2026. [2] P. Lewisetal.,NeurIPS, vol. 33,pp.9459-9474,2020. [3] Recentstudyonrelationship-awareretrieval,2024.\\n\\n[4]Recent study on efficient context linking, 2024. [4] C.Piech,J.Bassen,J.Huang,S.Ganguli,M.Sahami,L.J.Guibas, and J. Sohl-Dickstein, \"Deep Knowledge Tracing,\" Advances in Neural Information Processing Systems, vol. 28, 2015. [5] S.Zhang,X.Wang,Y.Ma,andD.Wang,\"An AdaptiveLearning Method Based on Concept Graph,\" Frontiers in Educational Research, vol. 6, no. 6, 2023. [6] BAAIFlagEmbeddingTeam,embeddingmodeldocumentation, 2023. [7] PostgreSQLvector-searchextensiondocumentation. [8] J.Shietal.,AppliedSciences,vol.15,no.16,2025. [9] Y.Zhouetal., AppliedSciences,vol.14, no.17, 2024. [10] A.C.Graesser,S.Lu,G.T.Jackson,H.H.Mitchell,M.Ventura, [11] Olney, and M. M. Louwerse, \"AutoTutor: A Tutor with Dialogue in Natural Language,\" Behavior Research Methods, Instruments,andComputers,vol.36,no.2,pp.180-193,2004. [12] M.VanLehn,\"TheRelativeEffectivenessofHumanTutoring, Intelligent Tutoring Systems, and Other Tutoring Systems,\" Educational Psychologist, vol. 46, no. 4, pp. 197-221, 2011.

Copyright

Copyright © 2026 Prof. Nagarathna C, Arun Kumar A R, Bharath M, Girish D R, Maadhu Avati. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.