PhD
Chair of Computational Linguistics at LMU
Centrum für Informations- und Sprachverarbeitung
Oettingenstraße 67, 1. OG, Flügel C
D-80538 München
Biosketch
Shengqiang is a first-year Ph.D. student at the Center for Information and Language Processing at LMU Munich. He is very fortunately co-advised by Prof. Hinrich Schütze and Prof. Barbara Plank. Previously, he obtained my master’s degree from Peking University in July 2021. Shengqiang spent a gap year as a research assistant advised by Wei Lu at SUTD, Singapore. He made several internships in industries such as Microsoft Research Asia and Baidu Search.
relAI Research
Towards Controllable and Consistent Text-Guided Visual Generation
My PhD research focuses on controllable text-guided visual generation across images and videos. The goal is to develop generative models that can synthesize and edit visual content according to natural language instructions while maintaining semantic faithfulness, spatial structure, identity, and temporal consistency.
I am particularly interested in understanding how text conditions modern generative models, including diffusion-based and autoregressive models. My work studies how model-internal signals such as attention maps, logits, latent features, and visual tokens can be used to improve controllability, structure preservation, and consistency.