Fig. 1: Basic information of the present study.

a In the present study, a delayed estimation task was adopted. Participants attempted to memorize four saturated colors and report them after a 1-s retention interval. The report for each color was done by tapping the corresponding white square and selecting the appropriate color on a color wheel. b The guidance neural network includes four layers, respectively with 8, 100, 100, and 196 nodes and fully connected throughout. c The QCE-VWM model underwent iterative refinement. This refinement process, involving the identification of individual mechanisms and the determination of their interrelationships, was guided by both observational clues and theoretical insights.