Guangyu Wang | Network Generative Models

I am an undergraduate student at Dongbei University of Finance and Economics. During this period, I serve as a research assistant to Zhaonan Wang at New York University. Starting this year, I will begin my PhD studies at HKUST (Guangzhou) as a member of the GIT Lab, led by Jiawei Wang, where I will pursue exciting research on network generative models.

I am passionate about all forms of computer technology. I have compiled Android 6.0 system source code from scratch and maintain a personal NAS server with 80TB of storage capacity.

News

RSS

Oct 8, 2026: Our work Addressing Spatial Indistinguishability in Spatiotemporal Prediction via Optimal Transport-Guided Masking was accepted by Pattern Recognition. I am the first author.
Oct 5, 2026: My earliest research work, A Dynamic Teacher-Student Gating Framework for Traffic Prediction, was accepted for publication in Enterprise Information Systems, an ABS 2-star journal. I am the first author; I began this project about two years ago.
Aug 21, 2026: Two collaborative papers were accepted to EMNLP 2026 Findings: BizSage: A Self-Evolving Multi-Agent Framework for Business Research with Efficient Knowledge Retrieval and Caught in the Story: Narrative Captivity in Multi-turn LLMs Conversation.
Aug 13, 2026: I was recognized as a Best Reviewer for the KDD AI4S Track.

Old News

Aug 11, 2026: Our work RiskTraf: Risk-Extrapolated Residual Learning for Multi-Variate Traffic Flow Prediction was accepted to CIKM 2026 as an oral presentation. I am the first author.
Jul 26, 2026: Our work Counterfactual Load Forecasting with LLM-Structured Events and Representation Learning was accepted by Applied Energy.
Jul 20, 2026: Our work Tactus: Sensor-Driven Adaptive Feedback for Home Hand Rehabilitation was accepted to the UbiComp/ISWC 2026 Posters Track. See you in Shanghai!
Jul 16, 2026: I received funding from the OpenRefinery STEM Research Program, with an award of $2,400 per year.
Jun 2, 2026: Our collaborative paper Bridging Land Use and Transport Planning: An AI-Enabled Decision Support System was accepted by Transportation Research Part A.
May 15, 2026: Our collaborative paper Interpretable Deep Learning for Traffic Prediction: Revealing the Role of Urban Morphology in Spatially Heterogeneous Cities was accepted by Journal of Transport Geography.
May 11, 2026: Our work Net-Ev²: A Generative Simulator for Network Event Evolution was accepted to the KDD 2026 Research Track. I am the first author.
May 1, 2026: Our collaborative paper Analyzing Environment-Transport Relationships in Metropolitan Areas: A Vulnerability-Based Geographic Framework and Prediction Model was accepted by Journal of Transport Geography.
Apr 21, 2026: Our collaborative paper From Urban Dynamics to Intelligence: A Spatiotemporal Framework for Cross-Modal Urban Crowd Flow Management was accepted by Cities.
Apr 7, 2026: Our work on LLM hallucinations, PRISM: Probing Reasoning, Instruction, and Source Memory in LLM Hallucinations, was accepted to the ACL 2026 Main Conference.
Mar 23, 2026: Our work Bridging Urban Theory and Artificial Intelligence: A Multi-Agent Recommendation System for Sustainable City Development was published in npj Urban Sustainability, exploring the future of LLMs and urban research.
Feb 1, 2026: My paper Climate-Adaptive Transportation Infrastructure: Cross-Regional Solutions for Urban Resilience and Emergency Response was published in Urban Climate.
Jan 10, 2026: Three of our papers were accepted to AAAI Workshops, marking an exciting opportunity for future research.
Nov 19, 2025: Our work Frequency as Identity: A Fourier Hypernetwork for Spatiotemporal Forecasting was published in Applied Soft Computing. I am the first author.

Research

Data Mining for Real-World Problems

Hybrid dynamical systems evolve continuously, leaving traces of their spatiotemporal properties. I study data-mining methods that learn from these traces, with a focus on network event evolution [KDD 2026], residual correction for traffic forecasting [CIKM 2026], and models of virality and causal inference [Ongoing]. My goal is to extract reliable signals from noisy and constantly changing data to understand real-world problems and support decision-making.

Greatest GNN in ST

I am fascinated by the elegant properties of graph neural networks. I study the role of GNNs in spatiotemporal systems [Pattern Recognition], investigate when simpler models can replace GNNs [Applied Soft Computing], and explore the problem of spatial indistinguishability [Ongoing]. This direction combines model analysis with practical forecasting methods to improve the accuracy, robustness, and interpretability of spatiotemporal models.

Fourier hypernetwork for spatiotemporal forecasting

Urban Data Management

I study the interactions among people, infrastructure, and cities through artificial intelligence and spatiotemporal modeling methods for urban data management. My work includes multimodal representation [Cities], interpretable prediction [Journal of Transport Geography], theory-driven decision support [npj Urban Sustainability, Transportation Research Part A], and climate adaptation [Urban Climate]. My goal is to connect machine learning with urban knowledge so that models can support transparent decision-making and transfer across real-world urban contexts.

Climate-adaptive transportation infrastructure framework

Potential Collaboration

If you are also looking for potential collaboration opportunities, feel free to reach out. I enjoy exploring future possibilities with collaborators across different fields and have had many positive collaboration experiences.

Get in touch