Jiatao Zheng 郑佳涛
Ph.D. Candidate in Control Science and Engineering
South China University of TechnologyAbout Me
I am a Ph.D. student at South China University of Technology. I am passionate about exploring cutting-edge technologies and dedicated to developing innovative solutions for real-world challenges. Currently, I focus on developing intelligent systems and innovative platforms, aiming to advance technology and improve the quality of human life.
I believe in the power of interdisciplinary research and open collaboration. My approach combines rigorous engineering principles with creative problem-solving, always seeking to bridge the gap between theoretical advances and practical applications. I am particularly interested in leveraging emerging technologies to address real-world challenges.
Beyond research, I am an advocate for scientific communication and knowledge sharing. I enjoy breaking down complex concepts into understandable pieces and believe that the best innovations emerge from collaborative environments. Outside the lab, you can find me exploring new technologies, enjoying sci-fi movies, or brainstorming the next breakthrough innovation. Feel free to reach out if you share similar interests or just want to discuss ideas - I'm always open to interesting conversations! 😊
Interests
Robotics
xxxxxx
Education
Ph.D. in Control Science and Engineering, 2023 -
Shien-Ming Wu School of Intelligent Engineering, SCUT
B.E., 2019 - 2023
South China University of Technology
News
- 2024.08. Check our new open-source project for evolving agents in Research and development (R&D) here; Online demo is available here. More news from Microsoft Research will be expected!!!
- 2024.07. I win the "Star of Tomorrow" award from Microsoft Research Asia, which is the highest honor for excellent internship.
- 2024.06. I am invited to the summer workshop of CUHK business school. More discussion from different perspectives will be expected, especially in the field of recommendation system and data mining.
- 2024.06. We release the dataset of Idea Reviewing for LLM agents. Check the details here and in projects section.
- 2024.06. I was invited to visit Soochow University, Nanjing University of Science and Technology, and Tongji University for discussion about the development and future of intellectual property.
- 2024.06. I was invited to the Fifth China Intellectual Property Innovation and Development Forum to give a speech, check the gallery for more details. Thank the great welcome from Prof. Jianwei Dang and etc.
- 2024.05. One paper abut multiview clustering by Grassmann manifold is accepted by TNNLS. Check the early access version here.
- 2024.04. Great pleasure to bridge and help Tec-Do become official advertising agency of Microsoft! Check the news here.
Projects
Stock Trading
A simple demo of using Time Series Prediction and Reinforcement Learning to trade stocks.
Experience
Investigate Over-reliance issues in LLM
CMU & UCLA
September 2024 - Present · Pittsburgh
Working with Prof. Anthony Chen from UCLA and Prof. Sherry Wu from CMU. Investigate the over-reliance issues on LLM by designed large-scale user study, with quantization, statistics methods, customized webs and browser extension.
One paper submitted to CSCW 2025.
Automatous R&D LLM Agent and Quantitative Finance Strategy
Microsoft Research Asia
January 2024 - August 2024 · Hanover
Working with Dr. Jian Bian, Microsoft Research Asia. Research on Automatic Research and Development (R&D) and Quantitative Finance Strategy. Imagine a world where the R&D process is fully automated, and the LLM can automatic generate analysis results of every proposed ideas and propose new ideas. Project we led is open-sourced in: Github/RD-Agent. I also won the highest honor "Star of Tomorrow" awarded by MSRA.
One paper accepted by ICLR 2024 AGI Workshop.
Explore Heterophily, Oversmoothing and Relations Modeling
Dartmouth College
December 2023 - October 2024 · Hanover
Advised by Prof. Yujun Yan and Prof. Dawei Zhou. Explore the heterophily and oversmoothing in graph neural networks, developing principles for graph-based ML models that are both expressive and generalizable.
One paper will be submitted to KDD 2025.
Multi-relational Geometric Information Interaction Research Group
Xi'an Jiaotong University
April 2023 - December 2023 · Xi'an
Advised by Prof. Danyang Wu. Research on the graph neural network, geometric representation learning and their applications for RNA, protein, and brain networks.
Two paper accepted by WWW 2024, two papers pending.
Data Mining and Machine Learning Engineer
Tec-Do
September 2022 - January 2024 · Guangzhou
Responsibilities for CTR/CVR prediction and Smart Bidding. Project of the Year awarded.
- Build CTR (Click-through rate) predication model through GBDT and GNN for the Tec-Do Ads DSP platform, defeating 3rd party (Cusper) SaaS services. Achieved ROI 1.4x from 0.7x.
- Use DCN implemented smart bidding module under constraint budget. 30% ROI improvement.
- Apply negative pseudo partial labels to sampling negative samples for recall model on item-user bipartite graph.
- Build workflow generating realistic picture for e-commerce based on diffusion model.
- Two patent granted, two patents published, and two patents pending.
Publications
May, 2024 · In IEEE Transactions on Neural Networks and Learning Systems
Cross-View Approximation on Grassmann Manifold for Multiview Clustering
Propose a cross-view approximation on Grassman manifold (CAGM) model to address inconsistencies within multiview adjacency matrices, feature matrices, and cross-view combinations from the two sources.
January, 2024 · In The Web Conference 2024
FinReport: Explainable Stock Earnings Forecasting via News Factor Analyzing Model
Propose an explainable stock earning framework via news factor analyzing model. Formalizing news into semantic graph and learn embedding of it as a more expressible factor. Stock earning foresting and explainable earning module are built via aggregating and utilizing news factor and numeric stock factor, achieving SOTA performance on A-stock dataset. A detailed report will be generated with the power of LLM.
January, 2024 · In The Web Conference 2024
Simple Multigraph Convolution Networks
Propose a simple but effective multigraph convolution networks, focusing on exacting edge-level and subgraph-level credible matrix to help cross-view interaction.
October, 2023
NP2L: Negative Pseudo Partial Labels Extraction for Graph Neural Networks
Propose an unsupervised framework to exact negative relationship between nodes, based on pseudo partial labels, and augment the original graph into a signed graph. Then signed graph neural networks are used for learning node embeddings, achieving state-of-the-art performance on both link prediction and node classification tasks.
Awards
Stars of Tomorrow
Highest honor for excellent internship
TaiHu Innovation Prize
Highest scholarship awarded by the Wuxi governments
National Scholarship
Highest scholarship awarded by the Chinese governments
First Price Scholarship
Highest scholarship awarded by the South China University of Technology
First Price Scholarship
Highest scholarship awarded by the EXCELLENCE GROUP
Intel Cup Undergraduate Electronic Design Contest
Second Prize in the Intel Cup Undergraduate Electronic Design Contest by the project of Multimodal Fitness Intelligent Guidance System
Star Volunteer
Volunteer title awarded by SCUT Youth Volunteer Guidance Center
Acknowledgements
I have been fortunate to work with many talented and dedicated individuals who have generously shared their time and expertise with me. I am grateful for their support and encouragement.
- Qinyi Zhou, Chang Liu · @Tsinghua University
- Yuxuan Ren, Haoyu Chi · @USTC
- Prof. Xiang 'Anthony' Chen · @UCLA
- Prof. Sherry Wu · @CMU
- Prof. Danyang Wu · @Xi'an Jiaotong University
- Prof. Yujun Yan · @Dartmouth College
- Prof. Jin Xu, Lin Shu, Shidang Xu, Yahui Jia · @South China University of Technology
- Dr. Guanyong Lu, Dr. Pinde Chen, Guofeng Huang, Shuhao Li · @Tec-Do
- Xiao Yang, Xu Yang, Weiqing Liu, Jiang Bian · @Microsoft
- Prof. Dawei Zhou · @Virginia Tech
- Zeqi Ye · @NanKai University
I also thanks friends in anime club, who always bring me joy and mental support.
Gallery
Enjoy both research and life
Contact
Reach me by email if you have any questions.