Yuanrong Qixing Shifts to Large Models with DeepSeek Experts

Yuanrong Qixing, a leader in autonomous driving technology, has announced a total pivot in its technical roadmap following the hiring of several core experts from the DeepSeek AI laboratory. The company is moving away from traditional rule-based systems and modular architectures toward an "End-to-End Large Model" approach for vehicle control.

The new strategy involves training a single, massive neural network that takes raw sensor data as input and directly outputs driving commands. This approach, inspired by the reasoning capabilities of LLMs, allows the vehicle to handle "edge cases" that are impossible to code manually, such as navigating a chaotic construction zone or interpreting ambiguous hand signals from traffic officers.

The influx of talent from DeepSeek is expected to accelerate the deployment of Yuanrong Qixing’s L4-level autonomous systems. By leveraging DeepSeek’s expertise in efficient model training and long-context reasoning, the company aims to become the first to achieve mass-market profitability for fully driverless vehicles in urban environments.