Dr. Tianyuan Teng

My long-term research goal is to understand how humans learn abstract representations of the world or tasks from fragmented experience, how reward and goals modulate this learning process, and how they are used in adaptive decision-making.

My previous work focused on the representation of uncertainty and unsupervised learning—specifically, on how humans apply low-dimensional probabilistic structures to incrementally compress high-dimensional density curves (Teng et al., 2023, 2025).

My current research explores how humans use abstract world/task representations for hierarchical planning (Teng et al., 2025), and investigates the interplay between abstraction, uncertainty, memory, and computational efficiency.