Xiao Chen

Ph.D. Candidate · The Hong Kong Polytechnic University

I am a Ph.D. candidate at The Hong Kong Polytechnic University, where I am fortunate to be advised by Professor Qing Li and Professor Zhaoxiang Zhang. I obtained my M.S. with honors from ZJU CS.

I have interned at Tencent, Squirrel AI, and Amazon previously, where I met thoughtful and insightful researchers from industry. These experiences made me more open to pursuing early-stage ideas with broad practical impact and building new concepts from the ground up.

Current research interests: Lifelong Learning for agents, Causal World Model, AI for BioScience.

Past research field: Model Fairness, Model Faithfulness, Agentic Recommender System, Off and On-policy Knowledge Distillation, Computer Vision, Medical AI

Email  /  Google Scholar  /  Twitter  /  Github

• [07/01/2026] I am currently on the job market, seeking postdoc positions. If you are interested in our work or think I might be a good fit, please feel free to reach out.
• [06/11/2025] Delighted to be recognized as an Outstanding Reviewer for KDD 2025 February Cycle.
• [05/15/2025] Our paper on Small Language Model-based Recommendations is accepted by ACL 2025, Stay tuned!
• [12/20/2024] Grateful to receive AAAI-25 Student Scholarship.
• [12/10/2024] Our paper on flexible reflection removal is accepted by AAAI2025.
• [05/01/2023] Join us in the “Trustworthy Recommender Systems” tutorial in The Web Conference 2023. See you in Austin!
• [01/25/2023] Our work on Fairly adaptive negative sampling is accepted by The Web Conference 2023.

During my PhD study, my research centers on Trustworthy Personalization. I am building the next generation of personalization systems that are both agentic and robust — capable of autonomously planning, reasoning, and acting on behalf of users while actively combating bias, unfaithfulness, and spurious correlations across data, models, and decisions. Rather than optimizing accuracy alone, I design systems that earn trust by being transparent, fair, and faithful by design, ensuring reliability and generalizability in real-world deployment. My work spans the following directions:

• Fairness: mitigating bias and ensuring equitable treatment across users and item groups, so personalized outcomes do not amplify existing disparities. [WWW'23]
• Reliable Edge Deployment: compressing and adapting personalization models for efficient, robust deployment on resource-constrained edge devices via off/on-policy distillation. [ACL'25] [under review]
• Faithful User Modeling: learning user representations that honestly reflect true preferences and intents, rather than spurious correlations or shortcut signals. [under review]
• Agentic System: building autonomous systems that perceive, reason, and act to serve user goals across different modalities. (i) Agentic Recommender System: empowering recommenders to act as autonomous agents that plan, reason, and interact to better satisfy user goals. [under review] (ii) Multimodal Interactive Image Restoration: building systems that leverage multimodal human interaction to guide and improve image restoration, enabling flexible and intuitive user control. [AAAI'25]
• Continual Learning for Personalization: enabling systems to learn continuously online once deployed — adapting to evolving interests and non-stationary environments without catastrophic forgetting. [under review]

• [TKDE 2026] A Low-Rank Perspective on Similarity Matrix Completion.
  Changyi Ma, Runsheng Yu, Xiao Chen, Youzhi Zhang, Zhen Lei.
• [ACL 2025 Findings] C2KD: Cross-Layer and Cross-Head Knowledge Distillation for Small Language Model-Based Recommendation.
  Xiao Chen, Changyi Ma, Wenqi Fan, Zhaoxiang Zhang, Qing Li.  [code]
• [AAAI 2025] FIRM: Flexible Interactive Reflection Removal.
  Xiao Chen, Xudong Jiang, Yunkang Tao, Zhen Lei, Qing Li, Chenyang Lei, Zhaoxiang Zhang.  [arXiv] [code]
• [WWW 2023] Fairly Adaptive Negative Sampling for Recommendations.
  Xiao Chen, Wenqi Fan, Jingfan Chen, Zhaoxiang Zhang, Qing Li.  [arXiv] [code]
• [arXiv 2023] A Comprehensive Survey on Trustworthy Recommender Systems.
  Wenqi Fan, Xiangyu Zhao, Xiao Chen, Jingtong Gao, Lin Wang, Qidong Liu, Yiqi Wang, Han Xu, Lei Chen, Qing Li. [arXiv]

• 2026: Internship at Nanyang Technological University (NTU), Singapore.
• 2025: Research Intern at Tencent, Shenzhen, working on Agentic Recommender Systems.
• 2022 – Present: Ph.D. in Computer Science at The Hong Kong Polytechnic University, advised by Chair Professor Qing Li and Professor Zhaoxiang Zhang.
• 2020 – 2022: Full-time AI Algorithm Engineer.
• 2020: Outstanding Graduates, Zhejiang Province
• 2017 – 2020: MSc in Computer Science and Technology at Zhejiang University, China, advised by Prof. Lanfen Lin and Prof. Ruofeng Tong.
• 2018 – 2019: Visiting Student at Ritsumeikan University, Japan, advised by Prof. Yen-Wei Chen.
• 2016: China National Scholarship
• 2016: First Prize in Asian Super Computer Competition
• 2013 – 2017: BSc in Software Engineering at Northwestern Polytechnical University, China.

I write paper-reading notes and literature surveys. Click a title to read the full post.

• Continual Learning Memory Management: A Survey 2026-07-01

  A comprehensive survey of memory management methods in continual learning, organized by techniques: experience replay, generative replay, gradient constraints, dual memory, replay-free approaches, and LLM-Agent methods.

• Recommendation as Generation (RaG): Paper Reading 2026-07-01

Tutorial Co-organizer
Trustworthy Recommender Systems: Foundations and Frontiers — KDD, WWW, IJCAI 2023

Conference & Journal Reviewer
NeurIPS (2024–2026), ICLR (2025–2026), ICML (2025),
KDD (2025–2026), WWW (2025),
ACL ARR (2026),
AAAI (2023, 2026), ECCV (2024), ACM MM (2024, 2026), AISTATS (2025–2026),
TOIS (2025), TKDD (2024–2025), TAI (2024)

Student Volunteer
ICDE 2025, ACL 2025

• ⏳ AI Conference Deadlines: Countdowns to top CV/NLP/ML/Robotics/AI conference deadlines.
• 📝 Writing Perfect Papers: A long talk on how to write perfect papers (by Prof. Baochun Li @UofT).
• 🍑 Lilian Weng's Blog: Deep insights for many ML problems.
• 📡 Scientific Spaces: ML|NLP theory blog with deep explorations.

I enjoy hiking, tennis, ukulele, hip-hop, and photography in my spare time.

Thanks for Jon Barron's template