Zhepeng Cen

zcen[at]andrew.cmu.edu

I am a fourth-year Ph.D. student in Safe AI Lab at Carnegie Mellon University, advised by Prof. Ding Zhao. I obtained my bachelor's degree at the Department of Automation, Tsinghua University. Prior to CMU, I had a wonderful time as a summer intern in USC INK Lab.

My research lies at the intersection of reinforcement learning and optimization. I am interested in how to facilitate the safety and data efficiency of the learning-based autonomy.

News

[2025/01] One paper on bridging the training-inference gap in LLM is accepted by TMLR.
[2024/09] I spent a wonderful summer as an intern in AWS! I worked with Rasool Fakoor and Yao Liu on bridging the training-inference gap in LLM.
[2024/09] One paper about data distribution shaping is accepted by NeurIPS 2024.
[2024/05] One paper on safe RL is accepted by ICML 2024. We achieved SOTA performance (02/2024) on safety-gymnasium. Check our codes for more details.
[2024/01] One paper on offline RL is accepted by ICLR 2024.
[2023/09] One paper on safe RL is accepted by NeurIPS 2023.
[2023/01] One paper is accepted by ICLR 2023.
[2022/05] One paper is accepted by ICML 2022.

Publications

(* indicates equal contribution)

Bridging the Training-Inference Gap in LLMs by Leveraging Self-Generated Tokens

Zhepeng Cen, Yao Liu, Siliang Zeng, Pratik Chaudhar, Huzefa Rangwala, George Karypis, Rasool Fakoor

TMLR; ICLR 2025 SSI-FM Workshop

Paper

OASIS: Conditional Distribution Shaping for Offline Safe Reinforcement Learning

Yihang Yao*, Zhepeng Cen*, Wenhao Ding, Haohong Lin, Shiqi Liu, Wenhao Yu, Tingnan Zhang, Ding Zhao

NeurIPS 2024

Paper / Website / Code

Feasibility Consistent Representation Learning for Safe Reinforcement Learning

Zhepeng Cen, Yihang Yao, Zuxin Liu, Ding Zhao

ICML 2024

Paper / Website / Code

Gradient Shaping for Multi-Constraint Safe Reinforcement Learning

Yihang Yao, Zuxin Liu, Zhepeng Cen, Peide Huang, Tingnan Zhang, Wenhao Yu, Ding Zhao

L4DC 2024

Paper / Website / Code

Learning from Sparse Offline Datasets via Conservative Density Estimation

Zhepeng Cen, Zuxin Liu, Zitong Wang, Yihang Yao, Henry Lam, Ding Zhao

ICLR 2024; ICML 2023 Frontiers4LCD Workshop

Paper / Code

Datasets and Benchmarks for Offline Safe Reinforcement Learning

Zuxin Liu*, Zijian Guo*, Haohong Lin, Yihang Yao, Jiacheng Zhu, Zhepeng Cen, Hanjiang Hu, Wenhao Yu, Tingnan Zhang, Jie Tan, Ding Zhao

DMLR 2024; RSS 2023 Safe Autonomy Workshop (Spotlight)

Paper / Website / Code (OSRL) / Code (DSRL) / Code (FSRL)

Constraint-Conditioned Policy Optimization for Versatile Safe Reinforcement Learning

Yihang Yao*, Zuxin Liu*, Zhepeng Cen, Jiacheng Zhu, Wenhao Yu, Tingnan Zhang, Ding Zhao

NeurIPS 2023; RSS 2023 Safe Autonomy Workshop

Paper / Code

Trustworthy Reinforcement Learning Against Intrinsic Vulnerabilities: Robustness, Safety, and Generalization

Mengdi Xu*, Zuxin Liu*, Peide Huang*, Wenhao Ding, Zhepeng Cen, Bo Li, Ding Zhao

Under review

Paper

On the Robustness of Safe Reinforcement Learning under Observational Perturbations

Zuxin Liu, Zijian Guo, Zhepeng Cen, Huan Zhang, Jie Tan, Bo Li and Ding Zhao.

ICLR 2023; ICML 2022 SL4AD Workshop (Best Paper Runner-up)

Paper / Website / Code

Constrained Variational Policy Optimization for Safe Reinforcement Learning

Zuxin Liu, Zhepeng Cen, Vladislav Isenbaev, Wei Liu, Steven Wu, Bo Li and Ding Zhao.

ICML 2022

Paper / Code

Will This Idea Spread Beyond Academia? Understanding Knowledge Transfer of Scientific Concepts across Text Corpora

Hancheng Cao*, Mengjie Cheng*, Zhepeng Cen*, Daniel A. McFarland, Xiang Ren

EMNLP (Findings) 2020

Paper

Academic Services

Conference Reviewer: ICML, NeurIPS (top reviewer@24), ICLR, CVPR, AISTATS.