Yichen Liu

Papers

Total Citations

H-Index

About

Yichen Liu is a researcher specializing in multi-modal learning and representation learning, with a particular focus on bridging the gap between different sensory modalities in machine learning systems. Their most notable contribution, "Improving Multi-Modal Learning with Uni-Modal Teachers" (2021), addresses a critical challenge in the field: the tendency of jointly trained multi-modal fusion models to underperform their uni-modal counterparts due to imbalanced learning dynamics. By leveraging uni-modal teachers to guide multi-modal learning, Liu's work introduces a principled framework that significantly improves the quality of learned representations — a development with meaningful implications for real-world robotic applications where reliable perception across modalities is essential. This paper has garnered 25 citations since its publication, reflecting its relevance to the growing community of researchers working on audio-visual learning, sensor fusion, and embodied AI. Liu's research sits at an important intersection of representation learning and robotics, offering practical solutions to longstanding optimization challenges in multi-modal systems. Their work is particularly valuable for students and researchers navigating the complexities of training models that must integrate and balance information from heterogeneous data sources.