Yizeng Han

Papers

Total Citations

H-Index

About

Yizeng Han is a rising researcher at the forefront of efficient deep learning and embodied AI, with a focus on making large-scale models practical for real-world deployment. His seminal survey, "Computation-efficient deep learning for computer vision: A survey" (2026, 19 citations), provides a comprehensive roadmap for reducing the computational burden of deep networks while maintaining high performance—a critical challenge as models grow in complexity. This work has become a foundational reference for researchers seeking to bridge the gap between state-of-the-art accuracy and real-time applications, particularly in autonomous systems. Han’s most notable contribution is the DeeR-VLA framework (2024), which introduces dynamic inference for Multimodal Large Language Models (MLLMs) in robotic execution. By enabling MLLMs to adaptively allocate computational resources based on task demands, DeeR-VLA significantly improves efficiency without sacrificing reasoning capabilities, addressing a key bottleneck in generalist robotics. This work has garnered attention for its potential to realize robots that understand complex human instructions and perform diverse embodied tasks. With a growing citation impact and a clear focus on practical AI, Han is shaping the future of computation-efficient, intelligent systems.