Original title: Quantum Sparse Autoencoders for Q-Matrix Estimation in Cognitive Diagnosis
Arif Hassan Zidan, Yi Pan, Bowen Guo, Xiang Li +4 · September 2026
Tests try to measure student abilities, but figuring out which exact skills each question requires is hard. Researchers use a table called a Q-matrix to map questions to skills. Building this table from student answers is tough when skills overlap or when real students do not act like tidy math models. To solve this, researchers built a quantum sparse autoencoder. It takes a student's binary answers of right and wrong, feeds them into a quantum circuit, shrinks them into a simplified pattern, and translates that pattern into a Q-matrix. When tested against a classical autoencoder, the classical system reached higher average accuracy in some simulations. However, the quantum method was much more reliable, showing less variance in 49 of 60 simulated tests. It also beat the classical model on 6 of 9 real-world tests.
Imagine you're trying to figure out which ingredients each recipe requires by watching people cook. Sometimes they skip steps or use substitutes, so it's hard to tell exactly what's needed. In this paper, estimating a Q-matrix is like figuring out the ingredient list for each dish from observing cooks, and the quantum method helps do this more consistently despite variations in cooking styles.
This could help educators create more reliable tests that accurately identify student skills, leading to better personalized learning for students and more effective teaching strategies for educators.
The quantum method is not universally more accurate than classical methods; it primarily offers better stability and performs well on some real datasets. Further validation may be needed.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
Quantum Sparse Autoencoders for Q-Matrix Estimation in Cognitive Diagnosis — Arif Hassan Zidan, Yi Pan, Bowen Guo et al.
Abstract licence: CC0
Explanation generated by Synthesis from the abstract — it is not the original paper. This is a preprint, so it hasn't necessarily been peer reviewed.