News
April 2026
Our paper DynamicPO received the Best Paper Award at DASFAA 2026 (the sole Best Paper Award winner selected from 834 valid Research Papers and 48 valid Industrial Papers).
Sep 2025
Started my Ph.D. studies, jointly trained by Shanghai Innovation Institute and USTC.
Aug 2025
Wrapped up my one-year internship at Meituan Ad Algorithm Team (Shanghai → Chengdu).
|
Xingyu Hu (胡兴宇)Ph.D. Student
Shanghai Innovation Institute (Joint Training) Shanghai, China
Email: |
I am a first-year Ph.D. student jointly trained by the Shanghai Innovation Institute (advised by Prof. Wenqi Shao 邵文琪) and the University of Science and Technology of China (advised by Prof. Xiaojun Chang 常晓军 at the Future Media Lab). My research focuses on agentic AI and preference optimization for large language models, with particular interest in LLM-based recommendation and long-horizon planning agents. I am the first author and co-corresponding author of DynamicPO, which received the Best Paper Award at DASFAA 2026 (the sole Best Paper Award winner selected from 834 valid Research Papers and 48 valid Industrial Papers). Prior to my Ph.D., I completed my master's-stage studies at USTC (2023-2025), advised by Prof. Xiangnan He (何向南) at the Lab for Data Science, and earned a Bachelor's degree from the Innovation Class of Computer Science at Hefei University of Technology (2019-2023). I also spent over a year as a research intern at Meituan Ad Algorithm Team.
Publications
(* indicates corresponding author.)| DASFAA 2026 Best Paper |
DynamicPO: Dynamic Preference Optimization for Recommendation
Xingyu Hu*, Kai Zhang, Jiancan Wu*, Shuli Wang, Chi Wang, Wenshuai Chen, Yinhua Zhu, Haitao Wang, Xingxing Wang, and Xiang Wang DASFAA 2026 (CCF-B; Accept Rate 27.3%) Best Paper Award (the sole Best Paper Award winner selected from 834 valid Research Papers and 48 valid Industrial Papers) |
Selected Projects
|
Agentic Tool Use
Trained tool-calling agent models that achieve leading open-source results on tool-use benchmarks including BFCL-v4 and Tau2-bench. |
|
Reflection-Augmented Retrieval-Augmented Generation
Designed a self-debate reflection mechanism that evaluates hallucination risk and the relevance of retrieved evidence, automatically generating sub-questions to supplement retrieval when information is insufficient. |
Awards & Honors
|
Apr 2026 DASFAA 2026 Best Paper Award (the sole Best Paper Award winner selected from 834 valid Research Papers and 48 valid Industrial Papers)
Dec 2024 Second Prize, BAAI Industry Data Application LLM Challenge
Oct 2024 First-class Academic Scholarship, University of Science and Technology of China
Apr 2024 Bronze-level (Stage 2), WorldQuant Quantitative Challenge 2024
|
Gallery
DASFAA 2026 Best Paper Award
Industry Experience
| Research Intern, Meituan (Chengdu), Ad Algorithm — Off-site Advertising Group, May 2025 - Aug 2025 Continued research on LLM-based recommendation and preference optimization, with a focus on real-world deployment in off-site advertising scenarios. |
| Research Intern, Meituan (Shanghai), Ad Algorithm — To-Home Search CTR/CVR Group, Aug 2024 - May 2025 Worked on preference optimization for LLM-based recommendation, frontier-tech follow-ups in recsys and LLMs, and participated in Meituan's internal AI Coding project. |
| Research Intern, Cognitive Recommendation Team, Data Space Research Institute, Hefei Comprehensive National Science Center, Mar 2024 - Jul 2024 Built LLM-based scoring of stock announcements, sub-concept extraction from financial news, and retrieval knowledge graph construction. |
Education
| Shanghai Innovation Institute & University of Science and Technology of China (Joint Training) Ph.D. in Artificial Intelligence / Information and Communication Engineering Sep 2025 - Present Advisors: Prof. Wenqi Shao (邵文琪, SII) and Prof. Xiaojun Chang (常晓军, USTC Future Media Lab) |
| University of Science and Technology of China (USTC) M.E. in Big Data Technology and Engineering Sep 2023 - Aug 2025, Hefei, China Advisor: Prof. Xiangnan He (何向南), Lab for Data Science |
| Hefei University of Technology (HFUT) B.E. in Computer Science and Technology (Innovation Class) Sep 2019 - Jun 2023, Hefei, China |
Last update: May 2026. Webpage template borrows from Prof. Xiangnan He.