Zixin Wen

Papers

Total Citations

H-Index

About

Zixin Wen is a researcher whose work lies at the intersection of multi-modal learning, representation learning, and the theoretical foundations of deep learning. Wen’s most notable contribution is the development of innovative training paradigms that improve how models integrate information from different modalities, such as vision and language. In their highly cited 2021 paper, “Improving Multi-Modal Learning with Uni-Modal Teachers,” Wen introduced a novel framework that leverages pre-trained uni-modal models to guide multi-modal fusion, effectively addressing the common problem of models failing to learn robust representations from all modalities during joint training. This work, which has garnered 25 citations, has provided a practical and principled solution for building more reliable multi-modal systems, particularly relevant for real-world robotic applications where sensor fusion is critical. Wen’s research is characterized by a deep understanding of optimization dynamics and a commitment to bridging theory and practice, making their contributions essential reading for anyone working on multi-modal AI, representation learning, or the development of more sample-efficient and robust deep learning models.