Zhepeng Cen

About

Zhepeng Cen is a rising researcher in artificial intelligence, with a primary focus on **safe reinforcement learning (RL)** — a critical subfield dedicated to training AI agents that can operate reliably in safety-critical environments. His major contributions lie in both theoretical foundations and practical benchmarking. In his highly cited 2022 paper, "Constrained Variational Policy Optimization for Safe Reinforcement Learning" (16 citations), Cen addressed the instability and lack of optimality guarantees that plague traditional primal-dual safe RL methods, proposing a novel variational framework that provides stronger theoretical grounding. Building on this, his 2023 work, "Datasets and Benchmarks for Offline Safe Reinforcement Learning" (5 citations), introduced a comprehensive benchmarking suite that includes standardized datasets and evaluation protocols, enabling the community to rigorously compare algorithms for learning safe policies from static data — a crucial step toward real-world deployment. By bridging theory and reproducible practice, Cen’s work is shaping how researchers develop and assess safety-aware agents, making him a notable voice in the growing field of trustworthy AI.

Research Focus

Key Achievements

Top Papers

1. 1
   Constrained Variational Policy Optimization for Safe Reinforcement Learning

   16 citations · 2022
2. 2
   Datasets and Benchmarks for Offline Safe Reinforcement Learning

   5 citations · 2023

Key Collaborators

Contact & Links