Li Hao
Ph.D. Candidate, School of Electronic Information, Wuhan University
I am Hao Li, a Ph.D. candidate at the School of Electronic Information, Wuhan University, advised by Professor Hao Jiang.
My research focuses on the intersection of social networks, graph neural networks, and large language models. I also collaborate with Dr. Yulia Gel and Dr. Yuzhou Chen on persistent homology for graph learning.
I am particularly interested in dynamic graph representation learning, opinion dynamics on graphs, and the use of large language models for social simulation and computational social science.
News
- 2026. 'Large Language Models as Topological Thinkers' was accepted to ICML 2026.
- 2026. 'Agentmandering' was accepted to AAAI 2026.
- 2025.05. 'TMetaNet' was accepted to ICML 2025.
- 2025.01. 'UniGO' was accepted to WWW 2025.
- 2024.12. 'Political Actor Agent' was accepted to AAAI 2025.
- 2024.05. 'Dynamic Neural Dowker Network' was accepted to KDD 2024.
Research
Topological Learning for Dynamic Graphs
I develop topology-aware representation learning for temporal, higher-order, and evolving relational data by combining persistent homology, Dowker and zigzag constructions, and graph neural networks. My recent work includes neural persistence-diagram approximation and topology-guided link prediction.
Social Networks and Opinion Dynamics
I study opinion propagation and social influence in complex networks. My work combines topic-specific social media data, synthetic opinion-dynamics datasets, and graph neural networks to build robust opinion-prediction models.
LLM Agents for Social Simulation
I build large language model agents for political and social simulation, with a focus on legislative decision prediction and fair redistricting. This work uses multi-source behavioral data and game-theoretic interactions among agents.
Selected Publications
-
Large Language Models as Topological Thinkers: A Benchmark on Graph Persistent Homology. International Conference on Machine Learning (ICML 2026) Paper -
Agentmandering: A Game-Theoretic Framework for Fair Redistricting via Large Language Model Agents. Proceedings of the AAAI Conference on Artificial Intelligence, 40(45): 38862-38870, 2026 Paper -
TMetaNet: Topological Meta-Learning Framework for Dynamic Link Prediction. International Conference on Machine Learning (ICML 2025) Paper -
UniGO: A Unified Graph Neural Network for Modeling Opinion Dynamics on Graphs. The Web Conference (WWW 2025) · Oral Presentation Paper
