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.
Previously, I studied math (BS) and CS/AI (MS) at Stanford where I was fortunate to work with many amazing mentors including Trevor Hastie, Chelsea Finn, and Emma Brunskill. I also worked concurrently as an economic consultant with Paul Milgrom at Auctionomics.
I'm broadly interested in RL, AI alignment, and 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。
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，有幸受教于 Trevor Hastie、 Chelsea Finn 与 Emma Brunskill 等多位老师。同期我还在 Auctionomics 担任经济顾问，与 Paul Milgrom 共事。
我的研究兴趣主要在强化学习、人工智能对齐与博弈论。
News近况
- Oct 20262026年10月New preprint: BIRD, a benchmark for iterative reward design.新预印本：BIRD，迭代式奖励设计的基准。
- Aug 20262026年8月Two workshop papers at RLC 2026 in Montreal.两篇 workshop 论文入选蒙特利尔 RLC 2026。
- Aug 20262026年8月Started Anthropic research fellowship.开始 Anthropic 研究员项目。
- Jul 20262026年7月New preprint: RENEW, on repairing model exploitation from preferences.新预印本：RENEW，用偏好修复世界模型的可利用性。
- May 20262026年5月New preprint: Imperfect World Models are Exploitable.新预印本：Imperfect World Models are Exploitable。
- Sep 20252025年9月Started the PhD in Edinburgh.开始在爱丁堡读博。
Research研究
- 2026A bird's-eye view of iterative reward designnew
- 2026RENEW: towards learning world models and repairing model exploitation from preferences
- 2026Imperfect world models are exploitable
- 2025Repairing reward functions with feedback to mitigate reward hacking
- 2025CompressedBeliefMDPs.jl: a Julia package for solving large POMDPs with belief compression
- 2025ExpFamilyPCA.jl: a Julia package for exponential family principal component analysis
- 2024Learning to explore in POMDPs with informational rewards
- 2023DynaDojo: an extensible platform for scaling analysis in dynamical system identification
Teaching教学
- 2024Algorithmic game theory course reader
- 2023Market design course reader