Precision Abacus‑based Representation Learning (PEARL)

We present the Precision Abacus-based Representation Learning (PEARL) dataset, an ultra-dense longitudinal resource tracking three adults over 280 days of Abacus-based Mental Calculation (AMC) learning. Each participant completed 52 to 55 MRI scans, yielding nearly 4,000 minutes of high-quality multi-echo multiband fMRI data (resting-state, task-based, and natural stimuli) alongside daily behavioral assessments. In the current study, AMC, as an ideal model for study learning-induced plasticity, transitioning from embodied tool use to abstract mental manipulation, provides both ecological validity and a clear staged learning progression.