Adaptive Knowledge Transfer for Cross-Disciplinary Cold-Start Knowledge Tracing

2025-11-25

AI Summary

Cross-disciplinary cold-start knowledge tracing (CDCKT) faces severe challenges due to extremely sparse interaction data in the target discipline and the absence of overlapping entities, rendering existing mapping-based approaches inadequate for modeling complex cross-disciplinary knowledge associations. To address this, we propose AdaptKT—a novel adaptive knowledge transfer framework integrating Mixture of Experts (MoE) and Generative Adversarial Networks (GAN). Its core innovation lies in leveraging source-discipline knowledge-state clusters as learnable cross-domain semantic bridges, enabling fine-grained feature alignment and disentanglement without shared entities. AdaptKT synergistically combines pretrained representations, clustering-guided MoE gating, and an adversarial discriminative module to significantly enhance knowledge-state modeling under few-shot target-discipline settings. Evaluated across 20 extreme cross-disciplinary cold-start scenarios, AdaptKT consistently outperforms state-of-the-art methods, demonstrating strong generalizability and practical applicability.

Technology Category

• Machine Learning: Mixture of Experts (MoE)
• Knowledge Representation and Reasoning: Knowledge Acquisition
• Cognitive Modeling & Cognitive Systems: Adaptive Behavior

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization

Abstract

Cross-Disciplinary Cold-start Knowledge Tracing (CDCKT) faces a critical challenge: insufficient student interaction data in the target discipline prevents effective knowledge state modeling and performance prediction. Existing cross-disciplinary methods rely on overlapping entities between disciplines for knowledge transfer through simple mapping functions, but suffer from two key limitations: (1) overlapping entities are scarce in real-world scenarios, and (2) simple mappings inadequately capture cross-disciplinary knowledge complexity. To overcome these challenges, we propose Mixed of Experts and Adversarial Generative Network-based Cross-disciplinary Cold-start Knowledge Tracing Framework. Our approach consists of three key components: First, we pre-train a source discipline model and cluster student knowledge states into K categories. Second, these cluster attributes guide a mixture-of-experts network through a gating mechanism, serving as a cross-domain mapping bridge. Third, an adversarial discriminator enforces feature separation by pulling same-attribute student features closer while pushing different-attribute features apart, effectively mitigating small-sample limitations. We validate our method's effectiveness across 20 extreme cross-disciplinary cold-start scenarios.

Problem

• Addresses insufficient student interaction data in cross-disciplinary knowledge tracing
• Overcomes limitations of scarce overlapping entities between disciplines
• Mitigates inadequate mapping functions for cross-disciplinary knowledge complexity

Innovation

• Clusters student knowledge states into categories
• Uses mixture-of-experts network with gating mechanism
• Employs adversarial discriminator for feature separation