Signal UNVaMP: Neural Knowledge Tracing with Variational Regularization of Latent Knowledge Dynamics

Summary

Carson J. Cook, Ahmed J. Zerouali, Anthony Schmidt, and colleagues published a paper on arXiv (cs.CY) on August 4, 2026, introducing the Unified Neural Variational Measurement of Proficiency (UNVaMP) architecture, a knowledge-tracing method that integrates observed student-item interactions with internal memory to produce evolving latent representations of student knowledge. These representations support accurate predictions of future responses while enabling explicit control over the smoothness of estimated learning trajectories. UNVaMP can be configured as either a purely neural model or a hybrid model that predicts responses through an interpretable measurement function over the latent space; the researchers show that the pure neural configuration (UNVaMP-MLP) achieves the strongest predictive performance among compared models on three of four datasets, while a hybrid configuration using a 1PL MIRT measurement function lags only slightly behind, indicating the predictive cost of interpretability is modest. Beyond predictive accuracy, UNVaMP provides a principled mechanism for controlling volatility when estimating student latent variables, quantification of uncertainty over knowledge-state estimates, and flexible input specification supporting heterogeneous student-item interaction features. Using an experimental dataset, the authors show auxiliary inputs induce structured changes in the hybrid model's predictive behavior, and a simulation study shows UNVaMP yields well-behaved knowledge-state estimates under controlled measurement conditions.