In the digital intelligence era, multimodal data (text, audio, images, video) has become essential for understanding user intentions and enabling data-driven decisions. However, semantic gaps and information conflicts across modalities often limit model performance. To address these challenges, Associate Professor Shuaiyong XIAO from the Department of Management Science and Engineering of School of Economics and Management, Tongji University (Tongji SEM) has advanced multimodal user behavior modeling and intelligent inference, with two collaborative papers accepted by top-tier journals.

The MIS Quarterly paper introduces a mixed-grained recommendation framework with hierarchical preference chains, enabling dynamic balance between item-level and category-level recommendations for more accurate user intention matching.

The INFORMS Journal on Computing study identifies negative multimodal interactions that degrade prediction. It proposes a regularized deep learning framework to strengthen complementary signals and suppress conflicting information across features, modalities, and samples.

Professor XIAO is an Associate Professor and PhD Supervisor at Tongji SEM. His research focuses on multimodal business intelligence, explainable deep learning, and large-model reasoning, with applications in short-video marketing. He has published in MIS Quarterly, INFORMS Journal on Computing, Information Systems Research, and other leading journals. He leads projects supported by the National Natural Science Foundation of China and the Ministry of Science and Technology, and received the Shanghai Super Postdoctoral Fellowship.