Organizations: Institutes of Physical Science and Information Technology, Anhui University, Hefei, China · School of Computer Science and Technology, East China Normal University, Shanghai, China · School of Computer Science and Technology, Anhui University, Hefei, China
Knowledge tracing (KT) predicts a student's next response from their history. New platforms often have question texts but no response logs for training, motivating zero-shot cross-domain KT. Existing transfer and cold-start methods use question semantics, but cross-platform semantic discrepancies limit transferability. We propose TIZSKT, which separates behavioral sequence modeling from domain-local historical evidence construction and uses text as an index rather than a directly transferable predictive representation. The Mastery-Dynamics Encoder learns correctness dynamics without question identifiers or text embeddings. Domain-Local Question Addressing builds fixed question addresses using a frozen text encoder and an unsupervised per-domain projection. The Text-Addressing Kernel uses these addresses to aggregate the current student's previous outcomes into a question-specific historical correctness summary. Cross-Segment Continuation retains access to earlier responses without extending the behavioral encoder's input sequence. Evidence-Volume Read-Out preserves the total kernel mass discarded by normalization and combines its log-compressed value with the behavioral state and historical correctness summary to predict the next response. All trainable parameters are learned from source domains; target inference uses public question texts and the current student's observed history without target-domain training. Leave-one-domain-out experiments on three public datasets show that TIZSKT achieves the highest mean AUC and accuracy among evaluated baselines on all target domains. Ablations confirm the contributions of text addressing and cross-segment continuation.
Figures & tables
Figure 1. Visualization of frozen text representations for all questions across the three domains. Points are reduced for visualization alone: a 256-dimensional PCA projection is fitted to the pooled question set, followed by t-distributed stochastic neighbor embedding.
Figure 2. TIZSKT consists of the Mastery-Dynamics Encoder, Domain-Local Question Addressing , a Text Addressing Kernel , Cross-Segment Continuation and Evidence-Volume Read-out.
Dataset
Questions
Students
Interactions
Correct
Test int.
MoocRadar
2,513
14,224
~0.9M
81.7%
178,600
NeuralPS
948
4,918
1.38M
53.0%
284,701
XES3G5M
7,651
18,066
5.55M
79.5%
1,105,018
Table 1. Datasets and held-out target test splits under the leave-one-domain-out assignment.
MoocRadar
NeuralPS
XES3G5M
Method
Embedding
AUC
ACC
AUC
ACC
AUC
ACC
AKT ( Ghosh et al., 2020 )
random
0.7814 ± 0.0005
0.8245 ± 0.0006
0.7207 ± 0.0007
0.6605 ± 0.0023
0.6943 ± 0.0078
0.7885 ± 0.0077
AKT ( Ghosh et al., 2020 )
Qwen3-8B
0.7804 ± 0.0007
0.8171 ± 0.0021
0.7200 ± 0.0005
0.6351 ± 0.0323
0.6918 ± 0.0023
0.7804 ± 0.0151
RobustKT ( Guo et al., 2025 )
random
0.7927 ± 0.0066
0.8259 ± 0.0011
0.7159 ± 0.0017
0.6439 ± 0.0074
0.6829 ± 0.0031
0.7843 ± 0.0026
RobustKT ( Guo et al., 2025 )
Qwen3-8B
0.8020 ± 0.0009
0.8273 ± 0.0018
0.7189 ± 0.0014
0.6535 ± 0.0054
0.6845 ± 0.0027
0.7928 ± 0.0081
stableKT ( Li et al., 2024 )
random
0.7823 ± 0.0004
0.8235 ± 0.0019
0.7203 ± 0.0004
0.6620 ± 0.0016
0.7001 ± 0.0012
0.7913 ± 0.0035
Table 2. Target-domain test AUC and ACC, mean ± sample standard deviation over five seeds. Bold: best per column.
Figure 3. Sensitivity of the zero-training addressing predictor, plotted as the deviation of target-domain AUC from each domain’s production value, one line per domain; exact recomputation, no seed variance. Dashed line and star: production value.
Figure 4. Target-domain test AUC of the full model and of three removal variants.
Knowledge Tracing (KT) infers a student's knowledge state from past interactions to predict future performance. Conventional Deep Learning (DL)-based KT models are typically tied to platform-specific identifiers and latent representations, making them hard to transfer and interpret. Large Language Model (LLM)-based methods can be either ungrounded under prompting or overly domain-dependent under fine-tuning. In addition, most existing KT methods are developed and evaluated under a same-distribution assumption. In real deployments, educational data often arise from heterogeneous platforms with substantial distribution shift, which often degrades generalization. To this end, we propose RAG-KT, a retrieval-augmented paradigm that frames cross-platform KT as reliable context constrained inference with LLMs. It builds a unified multi-source structured context with cross-source alignment via Question Group abstractions and retrieves complementary rich and reliable context for each prediction, enabling grounded prediction and interpretable diagnosis. Experiments on three public KT benchmarks demonstrate consistent gains in accuracy and robustness, including strong performance under cross-platform conditions.
Predicting future student responses to questions is particularly valuable for educational learning platforms where it enables effective interventions. One of the key approaches to do this has been through the use of knowledge tracing (KT) models. These are small, domain-specific, temporal models trained on student question-response data. KT models are optimised for high accuracy on specific educational domains and have fast inference and scalable deployments. The rise of Large Language Models (LLMs) motivates us to ask the following questions: (1) How well can LLMs perform at predicting students' future responses to questions? (2) Are LLMs scalable for this domain? (3) How do LLMs compare to KT models on this domain-specific task? In this paper, we compare multiple LLMs and KT models across predictive performance, deployment cost, and inference speed to answer the above questions. We show that KT models outperform LLMs with respect to accuracy and F1 scores on this domain-specific task. Further, we demonstrate that LLMs are orders of magnitude slower than KT models and cost orders of magnitude more to deploy. This highlights the importance of domain-specific models for education prediction tasks and the fact that current closed source LLMs should not be used as a universal solution for all tasks.
Prarthana Bhattacharyya, Joshua Mitton, Ralph Abboud +1
Knowledge Tracing (KT) models students' knowledge states based on learning interactions to predict performance. While deep learning-based KT models have boosted predictive accuracy, most models rely on deterministic vector embeddings and opaque latent state transitions, limiting interpretability regarding how specific past behaviors influence predictions. To address this limitation, we propose Probabilistic Logical Knowledge Tracing (PLKT), an interpretable KT framework that formulates prediction as a goal-conditioned evidence reasoning process over historical learning behaviors. Instead of representing knowledge states as deterministic vector embeddings, PLKT employs robust Beta-distributed probabilistic embeddings to represent student knowledge states. This probabilistic foundation allows us to model the uncertainty of historical behaviors and perform explicit logical operations (e.g., conjunction), constructing transparent reasoning paths that reveal how specific past interactions contribute to the prediction. Extensive experiments show that PLKT outperforms state-of-the-art KT methods while achieving superior interpretability. Our code is available at https://anonymous.4open.science/r/PLKT-D3CE/.
Siyu Wu, Cong Xu, Wei Zhang
1Shanghai Institute of AI Education, East China Normal University, Shanghai, 200241, China · 2Shool of Computer Sicence and Technology, East China Normal University, Shanghai, 200241, China