Jingfeng Yao

ReWorld: Learning Better Representations for World Action Models

Tianze Xia, Lijun Zhou, Kaixin Xiong, Jingfeng Yao, Yu Zhu, Zhenxin Zhu, et al.

World Action Models (WAMs) model future environment evolution under action conditioning, offering a scalable paradigm for autonomous driving. However, existing approaches focus largely on model architecture design, and how a WAM can efficiently learn better world representations…Read more