Bencheng Liao

Huazhong University of Science and Technology

Papers

Total Citations

H-Index

About

Bencheng Liao is an emerging researcher making significant strides in the intersection of generative AI and autonomous driving. His work centers on end-to-end autonomous driving systems, with a particular focus on harnessing cutting-edge diffusion models to advance robotic policy learning and vehicular decision-making. His most notable contribution, "DiffusionDrive" (2025), introduces a truncated diffusion model architecture for end-to-end autonomous driving, addressing a critical bottleneck in applying diffusion-based approaches to real-time robotic systems — namely, the computational burden of numerous denoising steps. By innovatively truncating the diffusion process, Liao demonstrates how multi-mode action distributions can be modeled efficiently without sacrificing the expressive power that makes diffusion models so compelling for complex policy learning tasks. Already accumulating 40 citations shortly after publication, this work signals strong community interest and positions Liao as a promising voice in the autonomous driving research landscape. His research bridges generative modeling and practical autonomous systems, offering solutions with genuine real-world applicability. Students and researchers working on autonomous vehicles, diffusion models, or embodied AI will find his contributions particularly relevant and forward-looking.