Human brains construct individualized global rankings from identical few-shot local learning

Ranking—a typical relational structure—helps people organize complex information and overcome cognitive load, yet in real-world settings it is often inferred from few-shot learning of partial comparisons. How the human brain makes relational inferences under such sparse conditions remains unknown. In a preregistered study combined with Magnetoencephalography (MEG) recordings, we show that even after identical few-shot local pair learning, individuals tend to construct distinct, stable global rankings that diverge from the actual ranking order. MEG recordings reveal that frontoparietal neural representations are reorganized to reflect each person's idiosyncratic ranking, distinct from innate ranking knowledge. These findings challenge the prevailing view that identical inputs yield shared outcomes across subjects and highlight the constructive and generative nature of human relational learning—human brains actively infer and impose structure under sparse data constraints.