Knowledge Tracing and Interpretability: A New Frontier

Knowledge Tracing (KT) is a crucial field for personalized education, aiming to model the evolution of students' knowledge to predict their future performance. Traditional deep learning models, while achieving high accuracy, often lack interpretability. Large Language Models (LLMs), on the other hand, offer strong reasoning capabilities but are limited by the context window size and the risk of hallucinations.

MERIT: An Innovative Approach

The MERIT (Memory-Enhanced Retrieval for Interpretable Knowledge Tracing) framework proposes a training-free solution that combines the reasoning capabilities of an LLM with structured pedagogical memory. Instead of updating model parameters, MERIT transforms interaction logs into an interpretable memory. This approach uses semantic denoising to categorize students into latent cognitive schemas and constructs a paradigm bank where representative error patterns are analyzed offline to generate Chain-of-Thought (CoT) rationales.

Inference and Advantages

During inference, a hierarchical routing mechanism retrieves relevant contexts, while a logic-augmented module applies semantic constraints to calibrate predictions. By grounding the LLM in interpretable memory, MERIT achieves state-of-the-art performance on real-world datasets without gradient updates. This reduces computational costs and supports dynamic knowledge updates, improving the accessibility and transparency of educational diagnosis.