Muzhi Wang

This paper introduces compressive learning, a framework that uses higher-order network structure to improve human knowledge acquisition by embedding node-degree inhomogeneity into pre-learning trajectories.

The paper develops a five-task EEG battery to detect dissociable neural biomarkers of reinforcement learning and motivation, identifying distinct neural markers across multiple tasks.

The paper shows humans create unique, self-consistent global rankings from limited local pairwise data, diverging from ground-truth, and reveals neural mechanisms underlying this process.

The paper shows that human working memory represents sequences using 2-D neural geometry, encoding item and chunk positions, and this structure persists across different experimental conditions.

The paper develops a computational model showing sustained and oscillated visual priming effects arise from attentional oscillation and temporal expectation, supported by reanalysis and new experiments.