News from the IT community on October 13th: Hello, Robotaxi announced yesterday the launch of the Generative Driving World Model HelloWorld. This model focuses on building driving data on demand and interactive strategy testing, integrating scene control, continuous generation, multi-sensor observation, and efficient reasoning capabilities to explore a specific path for world models to serve autonomous driving research and development.

According to the introduction, HelloWorld supports generating corresponding multi-view videos and lidar observations by describing requirements through natural language, or by directly editing the trajectories of vehicles, pedestrians, and moving objects. From inserting pedestrians crossing the road to adjusting the route of the vehicle in front, to switching between rain/snow and day/night conditions, researchers can construct scenarios based on specific needs, expanding on the long-tail situations that are difficult to fully cover in real-world data collection.

IT has learned that this model can also provide another scenario construction entry based on natural language. Users describe the situations they wish to generate, and the application process organizes the corresponding conditions and calls the model accordingly. Scenario descriptions such as construction, heavy traffic, pedestrian crossings, as well as rainy, snowy, foggy weather, and different lighting conditions, can all be used to guide the generation of content.