🎉 Our two RL / post-training papers, SLEA-RL and When Simulation Lies, were accepted to NeurIPS 2026.
Galleries of Prince 🤴🤴🤴
Ph.D. Researcher · Carnegie Mellon University
Prince Wang
Building capable, efficient, and safe LLM agents for human-centered collaboration.
- Human-Centric Agents
- Agent Post-Training
- Evaluation & Safety
Research Experience at
Hi, my name is Prince 👋 I am a PhD student 🎓 and entrepreneur 💼 in Information Systems at Carnegie Mellon University,
where I do research on GenAI and LLM agents 🤖. I am also the co-founder of WeMeow.ai (launching soon), an AI startup that builds the next generation of agentic-based social-media platform with proactive and personalized AI agents.
At CMU, my research focuses on building capable, efficient, and safe LLM agents:
(a) Human-Centric LLM Agents: Training LLM agents to perform human-centric tasks by learning from human feedback, preferences, and real-world interaction patterns (PRIME, AgentSocialBench);
(b) Cost-Efficient Reinforcement Learning: Developing cost-efficient RL methods for reasoning, multi-agent collaboration, and long-horizon agentic tasks (SLEA-RL);
(c) Building Robust Agents: Enhancing the safety and robustness of LLM agents through the design of environments and benchmarks that rigorously evaluate their behavior (AgentSocialBench).
I am looking for more research collaborations to build better LLM agents! We currently have several A100 GPU clusters capable of large-scale inference and RL post-training. If you are interested and have ideas, please reach out to us at princewang [at] cmu [dot] edu.
Education- 2022-present PhD in Information System, Carnegie Mellon University
- 2017-2022 5 year BS/MS in Computer Science, UC Santa Barbara
