Lunjun Zhang

Papers

Total Citations

H-Index

About

Lunjun Zhang is a rising researcher at the intersection of deep reinforcement learning, world modeling, and autonomous driving. His work centers on enabling intelligent agents to plan and reason about complex environments by learning structured representations of the world. In his highly cited paper "World Model as a Graph: Learning Latent Landmarks for Planning" (2020, 22 citations), Zhang introduced a novel framework that treats world models as graphs, allowing agents to decompose large-scale planning problems into interrelated subproblems—a hallmark of human-like intelligence. This approach marked a significant step forward in making reinforcement learning more efficient for long-horizon tasks. More recently, Zhang co-authored "Copilot4D: Learning Unsupervised World Models for Autonomous Driving via Discrete Diffusion" (2023, 7 citations), which scales world model learning to real-world robotic applications. By applying discrete diffusion processes, this work pushes the boundaries of unsupervised sequence modeling for autonomous driving, addressing a critical gap where progress has lagged behind language models. Zhang’s contributions are shaping how machines learn to navigate and plan in dynamic, unstructured environments, with growing impact in both academic and applied AI communities.