For Prospective Students

I mainly work on the principled design and theoretical analysis of efficient learning and optimization algorithms, with a particular focus on online learning, optimization, and reinforcement learning.

More recently, I have been extending these ideas to efficient inference and training of large language models. My current interests include speculative decoding, sparse and linear attention, efficient optimization for agentic systems, etc. Broadly, I am interested in developing principled algorithms that improve both the statistical and computational efficiency of modern AI systems.

Please refer to this publication page for my most recent papers.

If you are an NJU undergraduate student (AI, CS, SE, Math, Statistics, Physics, etc), I welcome self-motivated students to join my research projects.

Please note that successful undergraduate research requires a significant time commitment, including reading literature, deriving proofs, and implementing code. Therefore, an important prerequisite is to have sufficient capacity beyond regular coursework (“学有余力”). As a reference, this may correspond to a top 10% ranking, or around top 20% for students from the AI School; exceptions may be made for students with strong research experience or other notable achievements.

If you are passionate about the research and and can dedicate sufficient time, you are welcome to drop me an email with your CV, transcript (with ranking info), a brief description of your research interests, and any other relevant materials.

您可以参考此（有些过期的）宣讲幻灯片了解更多组内情况 [2025年本科生进组宣讲-v1011.PDF]，以及参考 [以往优秀进组本科生]

TIPS: I would also encourage you to take my course (Advanced Optimization 2026 Fall), which will provide useful technical tools for conducting research in my group, and you can also have a better understanding of my research.

TIPS: I encourage undergraduates to participate in competitions such as ASC and ICPC (FYI: i don't think mathematical modeling contest is useful), and to contribute to open-source projects such as vLLM. These experiences will strongly strengthen your engineering skills and build a solid analytical foundation for research.

I am looking forward to self-motivated students with strong math background or programming experience. Usually, I will take about 3–5 students each year. If you are interested in working with me, feel free to send me an email with your CV, transcript, a description of your research interests, and other related materials. Please start your email by stating that you have read this instruction.

您可以参考此FAQ文档了解更多本年度招生情况 [2027招生FAQ]

Note for master/PhD students: Unfortunately, I am unable to respond to most inquiries regarding the admission possibility for graduate students in my group. Admission is mostly handled at a department level in Nanjing University.

Note for undergraduate students: I'm happy to write recommendation letters for undergraduate students applying to PhD programs at top-tier universities. For detailed and personalized letters, you should have been part of my research group for at least a year, allowing me to get to know you well. Otherwise, I can only write as a course instructor, provided you have made a strong impression in the class.