Fig. 9: A thorough comparison of all models.

The QCE-VWM model (represented by a green bar), along with the 17 alternative models (represented by blue bars), the guidance neural network and its four reduced versions (represented by red bars), and the eight models in the factorial comparison analysis (represented by yellow bars), are all sorted based on their data fitting (NLLsr). Please note that uneven scales are used across different ranges because the fittings are densely clustered within two specific ranges: one for the QCE-VWM and alternative models, and another for six of the eight models in the factorial comparison analysis. Consequently, uneven scales are employed to highlight the distinctions within these two clusters. The QCE-VWM and alternative models greatly outperform those in the factorial comparison analysis due to their effective integration of more mechanisms. Furthermore, the QCE-VWM and several alternative models surpass the performance of the guidance neural network, presumably because the latter does not capture the underlying mechanisms as precisely as the former.