RESEARCH PAPER

Drive-HWM: Hierarchical World Models for Dynamic-Latent Guided Autonomous Driving

Zhaoxin Fan; Tianbao Zhang; Wenjun Wu; Xiaofeng Wang; Yeying Jin; Jian Zhao; Zheng Zhu; Shuicheng Yan

Classification

View four quadrants
Major category
WAMs
Quadrant
Not assigned
Architecture
Not assigned
Prediction paradigm
Not assigned
Source review status
Not assigned

Category review. A slow future-motion predictor provides dynamic latents to an observation-grounded driving action model; predictive latents remain active in action generation at deployment. Reading evidence

AT A GLANCE

Contribution

Drive-HWM couples a periodically refreshed predictor of future motion latents with an observation-grounded autoregressive driving policy. Optical flow supervises the slow branch; next-frame RGB tokens supervise the fast branch during training. NAVSIM tables report stronger aggregate driving scores, but conflicting prose and incomplete implementation details limit precise reproduction.

Abstract

An abstract has not been added yet.

Affiliations

School of Artificial Intelligence, Beihang University, Beijing, China; Shanghai Jiao Tong University, Shanghai, China; Dim12 AI; GigaAI; National University of Singapore, Singapore; TeleAI