A novel framework named DUal-STream diffusion (DUST) has been introduced to advance vision-language-action models by incorporating world-models. This approach is designed to enhance robotic policy learning, addressing the specific challenge of predicting next-state observations alongside action sequences. By effectively managing these prediction tasks, DUST aims to improve the integration of visual and linguistic inputs with action planning. The framework's application area centers on robotics, where accurate anticipation of future states is critical for decision-making. Recent connected coverage highlights the framework’s focus on overcoming prediction difficulties inherent in robotic control systems. This development reflects ongoing efforts to create more robust and adaptive models that can better understand and interact with complex environments. Overall, DUST represents a significant step toward more sophisticated vision-language-action integration in robotics.