Chi Jin (金驰)

My research focuses on the decision-making aspects of machine learning. We aims to develop intelligent agents capable of complex strategy, advanced reasoning, and planning. In the past, my group has primarily worked on establishing the theoretical foundations of machine learning, including areas such as reinforcement learning, multi-agent learning, game theory, statistical learning, and optimization. Recently, we have expanded our interests to include improving the reasoning abilities of LLM and developing LLM agents for tasks such as mathematics, coding, and complex games.

*I am recruiting undergraduate research interns, PhD students, and Postdoctoral researchers*. To apply, please email me with your CV attached. PhD applicants should also mention my name as a faculty member of interest in their statement of purpose. Due to the high volume of inquiries, I may not be able to respond until the hiring process begins. 

News

• I'm very excited to announce our project on LLM automatic theorem proving---Goedel-prover, which achieves SOTA performance on miniF2F,  PutnamBench, and Lean-workbook!

• I have been recently giving talks on beyond equilibrium learning in game theory. See also my Princeton course on foundations of reinforcement learning here, and my tutorial on multiagent reinforcement learning at Simons institute here.

Awards

• Sloan Research Fellowship, 2024

• E. Lawrence Keyes, Jr./Emerson Electric Co. Faculty Advancement Award, 2023

• NSF CAREER Award, Division of Information and Intelligent Systems, 2023

• Princeton Commendation for Outstanding Teaching for ECE 524, 2022 & 2024

• Princeton Commendation for Outstanding Teaching for ECE 539, 2021 & 2023

• Princeton School of Engineering and Applied Science (SEAS) Innovation Award, 2022

• Best Paper Award in ICLR 2022 workshop “Gamification and Multiagent Solutions”.

• Best Paper Award in ICML 2018 workshop “Exploration in Reinforcement Learning”.

Selected Paper

Goedel-Prover: A Frontier Model for Open-Source Automated Theorem Proving [arXiv]

• Yong Lin, Shange Tang, Bohan Lyu, Jiayun Wu, Hongzhou Lin, Kaiyu Yang, Jia Li, Mengzhou Xia, Danqi Chen, Sanjeev Arora, Chi Jin

• ArXiv Preprint

Securing Equal Share: A Principled Approach for Learning Multiplayer Symmetric Games [arXiv]

• Jiawei Ge*, Yuanhao Wang*, Wenzhe Li, Chi Jin

• ArXiv Preprint

Maximum Likelihood Estimation is All You Need for Well-Specified Covariate Shift [arXiv]

• Jiawei Ge*, Shange Tang*, Jianqing Fan, Cong Ma, Chi Jin

• International Conference on Learning Representations (ICLR) 2024

Optimistic MLE -- A Generic Model-based Algorithm for Partially Observable Sequential Decision Making [arXiv]

• Qinghua Liu, Praneeth Netrapalli, Csaba Szepesvari, Chi Jin

• Symposium on Theory of Computing (STOC) 2023

When Is Partially Observable Reinforcement Learning Not Scary? [arXiv]

• Qinghua Liu, Alan Chung, Csaba Szepesvári, Chi Jin

• Conference of Learning Theory (COLT) 2022.

Near-Optimal Learning of Extensive-Form Games with Imperfect Information. [arXiv]

• (α-β order) Yu Bai, Chi Jin, Song Mei, Tiancheng Yu

• International Conference on Machine Learning (ICML) 2022.

V-Learning -- A Simple, Efficient, Decentralized Algorithm for Multiagent RL [arXiv]

• (α-β order) Chi Jin, Qinghua Liu, Yuanhao Wang, Tiancheng Yu

• Mathematics of Operation Research (MOR) 2023

• Best Paper in ICLR 2022 workshop “Gamification and Multiagent Solutions”

Bellman Eluder Dimension: New Rich Classes of RL Problems, and Sample-Efficient Algorithms [arXiv]

• (α-β order) Chi Jin, Qinghua Liu, Sobhan Miryoosefi 

• Neural Information Processing Systems (NIPS) 2021.

Near-Optimal Algorithms for Minimax Optimization [arXiv]

• Tianyi Lin, Chi Jin, Michael. I. Jordan

• Conference of Learning Theory (COLT) 2020

Provably Efficient Reinforcement Learning with Linear Function Approximation [arXiv]

• Chi Jin, Zhuoran Yang, Zhaoran Wang, Michael I. Jordan

• Conference of Learning Theory (COLT) 2020

Is Q-learning Provably Efficient? [arXiv]

• Chi Jin*, Zeyuan Allen-Zhu*, Sebastien Bubeck, Michael I. Jordan

• Neural Information Processing Systems (NIPS) 2018. 

• Best Paper in ICML 2018 workshop "Exploration in RL"

Accelerated Gradient Descent Escapes Saddle Points Faster than Gradient Descent [arXiv]

• Chi Jin, Praneeth Netrapalli, Michael I. Jordan

• Conference of Learning Theory (COLT) 2018

How to Escape Saddle Points Efficiently [arXiv] [blog]

• Chi Jin, Rong Ge, Praneeth Netrapalli, Sham M. Kakade, Michael I. Jordan

• International Conference on Machine Learning (ICML) 2017.