Peking University · CS PhD Candidate

Hongcheng Wang

I am a fourth-year CS PhD candidate at Peking University. My supervisor is Prof. Hao Dong.

我预计于 2027 年 6 月博士毕业，目前也在关注具身大脑 / 具身智能 VLA、VLM / LLM 后训练等方向的研究与算法岗位机会，欢迎交流。

My current research interests are organized into two parts:

• Chain-of-thought Generation: I study reinforcement-learning post-training for large models, especially how to stably train <think> to better guide <answer>.
• Human-centered Robot Decision Making: I take a Cognitive Behavioral Theory's perspective to enable robots to understand human cognition (e.g., latent demands and preferences) and behavior (e.g., observable habits and norms), so they can better serve people.
• LLM-based Multi-Agent Reinforcement Learning (On going): I study reinforcement learning where LLMs act as multiple agents, with a particular focus on agent-level credit assignment and communication credit assignment.