An Empirical Investigation of Adaptive Cognitive Guidance Strategy Optimization Driven by Hierarchical Chain‐of‐Thought Reasoning in Large Language Models

ABSTRACT As large‐scale language models (LLMs) gain traction in education, the core challenge for intelligent tutoring has shifted from static knowledge transfer to dynamic cognitive companionship. This study proposes a novel adaptive cognitive tutoring strategy driven by a hierarchical chain‐of‐reasoning. The strategy constructs a three‐layer reasoning framework—strategic, tactical, and detailed—allowing complex problems to be decomposed into manageable cognitive units so learners can progressively deepen their understanding and problem‐solving. This study also designs a cognitive‐state diagnosis module that extracts and quantifies learners' multi‐dimensional features in real‐time, covering four key dimensions: knowledge mastery, error patterns, metacognitive status, and learning stage. Based on these precise diagnoses, the system dynamically adjusts the reasoning‐chain granularity and instructional tactics to provide individualized cognitive scaffolding. This study conducts a systematic empirical study on the MathDial math tutoring dialogue dataset. Results show that Hierarchical Chain‐of‐Thought–driven Adaptive Cognitive Tutoring Strategy (HCoT‐ACTS) achieves an 81.6% final‐problem resolution rate, a 9.5 percentage‐point improvement over the best baseline, and an average attainment in only 5.6 dialogue turns, demonstrating high efficiency and accuracy. Ablation studies confirm that the cognitive‐state diagnosis and dynamic adjustment of the reasoning chain are core contributors to the strategy's effectiveness. To evaluate generalizability, the study further validates the approach on the MathQA dataset across datasets and model settings. HCoT‐ACTS maintains significant advantages in new data and model environments, demonstrating strong adaptability and robustness across scenarios. In summary, this work offers an interpretable and operational technical framework for adaptive intelligent tutoring, provides solid empirical evidence for human–AI educational collaboration, and points to new directions for the intelligent future of education.