I'm an AI PhD student at the University of Edinburgh, co-advised by Ram Ramamoorthy (Edinburgh) and Mykel Kochenderfer (Stanford). I'm currently an Anthropic Research Fellow. My recent work is on reward hacking and model exploitation in RL.
Previously, I studied math (BS) and CS/AI (MS) at Stanford, where I worked with Trevor Hastie, Chelsea Finn, and Emma Brunskill. I also worked as an economic consultant with Nobel laureate Paul Milgrom at Auctionomics.
I'm broadly interested in alignment and physical AI (RL, robotics, game theory).
我是爱丁堡大学人工智能方向的博士生，由 Ram Ramamoorthy（爱丁堡）与 Mykel Kochenderfer（斯坦福）联合指导，目前也是 Anthropic Research Fellow。我近期的工作集中在强化学习中的奖励黑客（reward hacking）与模型利用（model exploitation）问题。
此前我在斯坦福取得数学学士与计算机科学／人工智能硕士学位，期间曾与 Trevor Hastie、 Chelsea Finn 与 Emma Brunskill 合作。我还在 Auctionomics 担任经济顾问，与诺贝尔经济学奖得主 Paul Milgrom 共事。
我的研究兴趣主要在人工智能对齐与物理AI（强化学习、机器人学、博弈论）。
Research研究
- A Bird's-Eye View of Iterative Reward Design
- Imperfect World Models are Exploitable
- RENEW: Towards Learning World Models and Repairing Model Exploitation from Preferences
- Repairing Reward Functions with Feedback to Mitigate Reward Hacking
- Learning to Explore in POMDPs with Informational Rewards
- DynaDojo: An Extensible Platform for Benchmarking Scaling in Dynamical System Identification
