Interpretable machine learning for thermoelectric materials design with Kolmogorov–Arnold networks

The discovery of high-performance thermoelectric materials requires models that are both accurate and interpretable. Traditional machine learning approaches, while effective at property prediction, often act as black boxes and provide limited physical insight. In this work, we introduce Kolmogorov–Arnold Networks (KANs) for the prediction of thermoelectric properties, focusing on the Seebeck coefficient and band gap. Compared to multilayer perceptrons (MLPs), KANs achieve comparable predictive accuracy while offering explicit symbolic representations of structure-property relationships. This dual capability enables both reliable predictions and physically interpretable functional forms, providing insight into the governing mechanisms of thermoelectric behaviour. Benchmarking against literature baselines highlights their robustness and generalisability, demonstrating that KANs constitute a practical framework for reverse engineering materials with targeted thermoelectric performance and bridging the gap between predictive power and scientific interpretability.