cv
General Information
| Name | Tianhao Wu |
| Born In | China |
Education
-
2021 - Now Ph.D.
University of California, Berkeley Electrical Engineering and Computer Sciences - Research on LLM alignment and reasoning via reinforcement learning and self-play.
-
2017 - 2021 Undergrad
Peking University Mathematics - I worked on deep learning theory and reinforcement learning theory.
Experience
-
2025 Summer Hudson River Trading (HRT)
- Quantitative research intern.
-
2024 May Meta
- Research on AI self-improving algorithms.
-
2024 Feb Nexusflow
- We trained Starling-7B-LM, a small language model outperforming Mixtral-7x8b and Gemini-Pro.
-
2023 Sep TikTok
- Applied RL to improve safety and robustness of recommendation system.
Honors and Awards
-
2015 - Gold Medal in Chinese Mathematics Olympiad (CMO)
-
2015 - Gold Medal in Russian Mathematics Olympiad (RMO)
Academic Interests
-
Agent swarms and self-improving agents
- Building systems where agents share skills and evolve from each other's experience.
-
RL for LLMs
- Improving LLM reasoning and instruction following via reinforcement learning and self-play.