Kai Yan

Ph.D. Student

Department of Computer Science

University of Illinois Urbana Champaign

Office: CSL 130, 1308 W Main St, Urbana, IL 61801, United States

Email: kaiyan3 [at] illinois [dot] edu

"We choose to go to the Moon in this decade and do the other things, not because they're easy, but because they're hard."
-- John F. Kennedy, 1962

Who am I?

Hi there! I am Kai Yan (颜开 in Chinese), a second-year Ph.D. student in the Department of Computer Science at University of Illinois Urbana-Champaign (UIUC), co-advised by Prof. Alexander Schwing and Prof. Yu-Xiong Wang. Prior to that, I get my Bachelor of Science degree in computer science at Peking University with a Summa Cum Laude and a national scholarship. In my high school years, I learned informatics and got a silver award in the National Olympiad of Informatics (NOI 2016).

My research interest is deep learning for better decision making, which is mainly deep reinforcement learning and imitation learning. I have conducted research in the following fields: 1) optimization with prediction, 2) multi-agent reinforcement learning, 3) demonstration-guided reinforcement learning and imitation learning, 4) decision transformer, and 5) Large Language Model (LLM)+MCTS.

You can check my CV, Github and Linkedin here. Don’t forget to check your daily tips at the top of this page!

Publications

Kai Yan, Alexander G. Schwing and Yuxiong Wang. Reinforcement Learning Gradients as Vitamin for Online Finetuning Decision Transformers. In NeurIPS, 2024. (Spotlight)[PDF][Website]

Andy Zhou, Kai Yan, Michal Shlapentokh-Rothman, Haohan Wang, Yuxiong Wang. Language Agent Tree Search Unifies Reasoning Acting and Planning in Language Models, In ICML, 2024. (Github >630 Stars)[PDF][Website][LangChain link][Invited Talk]

Kai Yan, Alexander G. Schwing and Yuxiong Wang. Offline Imitation from Observation via Primal Wasserstein State Occupancy Matching. In ICML, 2024. [PDF][Website]

Kai Yan, Alexander G. Schwing and Yuxiong Wang. A Simple Solution for Offline Imitation from Observations and Examples with Possibly Incomplete Trajectories. In NeurIPS, 2023. [PDF][Website]

Kai Yan, Alexander G. Schwing and Yuxiong Wang. CEIP: Combining Explicit and Implicit Priors for Reinforcement Learning with Demonstrations. In NeurIPS, 2022. [PDF][Website]

Kai Yan, Jie Yan, Chuan Luo, Liting Chen, Qingwei Lin and Dongmei Zhang. A Surrogate Objective Framework for Prediction+Optimization with Soft Constraints. In NeurIPS, 2021. [PDF]

Preprints

Kai Yan*, Zhenggang Tang*, Liting Sun, Wei Zhan, Changliu Liu. A Microscopic Pandemic Simulator for Pandemic Prediction Using Scalable Million-Agent Reinforcement Learning. arXiv:2108.06589, 2021.[PDF]

Kai Yan*, Yunlong Lu*. Algorithms in Multi-Agent Systems: A Holistic Perspective from Reinforcement Learning and Game Theory. arXiv:2001.06487, 2020. [PDF]

* Equal Contributions.

Working Experiences

I am working in ByteDance US mentored by Dr. JieCao Chen since Jun. 2024. I also worked in Microsoft Research Asia mentored by Dr. Jie Yan and Chuan Luo from Sept. 2020 to Jun. 2021, and was awarded Stars of Tomorrow.

Teaching Experiences

I was a teaching assistant for Computational Photography (Spring 2024) at UIUC, during which I held office hours, graded homework and designed exam questions. Prior to that, I was a teaching assistant for Introduction to Computer Systems (Fall 2019) at Peking University, where I led a seminar of 15 people every week, revising and expaning lessons taught in class, grading homeworks, and teaching how to write lecture notes.

Photos with Students and Teacher in PKU

Other Experiences

I served for the department of outreach in the EECS Student Union of Peking University in 2017-2018, and as the vice president of the EECS Student Union in the academic year of 2018-2019. We have organized awesome events with hundreds of participants, such as Hackathon and Freshmen Ball.

Photos of Our Hackathon

Hey, you find my easter egg! I am a lover of strategic PC games, such as Frostpunk, Civilization, Stellaris, Hearts of Iron, Victoria and Total Wars. That’s also something where you need to make good decisions :)