1.Causal Alignment for Large Language Models (2023.05–)
We propose a novel approach based on causal abstraction theory and mixture of experts to find the neural representation corresponding to a given causal variable in the causal structure model.
Portfolio
2.Social Resilience Research Based on Causal Discovery (2023.02–2023.05)
We characterize social resilience using the resilience of social cognitive networks generated by causal discovery. 
3.Modeling Information Cascading Behavior Based on Causal Inference (2022.09–2023.02)
We employ causal inference to quantify the extent of influence that internal motivation and external environment have on individual decision-making. By delving into the reasons behind individual behavior, it aims to furnish a solid foundation for the formulation of effective policies.
4.Deep Causal Learning: Representation, Discovery and Inference (2021.11–2022.09)
We introduce the latest advancements in three fundamental areas of deep causal learning: causal representation learning, deep causal discovery, and deep causal inference. 
5.Credible Influence Analysis in Mass Media Using Causal Inference (2021.06–2021.11)
Based on large-scale news media data, we use the cross convergent mapping method to construct a causal network of influence among countries, and superimposes the influence network generated by the interaction degree to form a credible influence network. 