RLVR-World: Training World Models with Reinforcement Learning

Jialong Wu, Shaofeng Yin, Ningya Feng +1

World models predict state transitions in response to actions and are increasingly developed across diverse modalities. However, standard training objectives such as maximum likeli…