Yingru LI
Member of Technical Staff
xAI
About me
I am a Member of Technical Staff at xAI. I earned my Ph.D. in Computer Science in 2025 from The Chinese University of Hong Kong (CUHK), where I had the privilege of being advised by Prof. Zhi-Quan (Tom) Luo, with Prof. Benjamin Van Roy on my thesis committee.
Prior to xAI, I was a Research Scientist at ByteDance. I also had the valuable opportunity to collaborate with Prof. Tong Zhang and Prof. John Hopcroft.
Research Vision
My research aims to develop intelligent agents capable of reliably interacting with complex environments. By bridging foundational theory with scalable algorithms, I advance reinforcement learning, large scale optimization, and large language model (LLM) reasoning to create systems for trustworthy decision-making.
🐦 Follow me on X for updates.
At xAI
- Science of RL (Grok 4.2) — identified and resolved critical training instabilities that benefit every stage of RL training across the entire pipeline, directly enabling the longest stable RL run to date
- Long-Horizon RL (Grok Next) — leading the development of recipes for higher intelligence per token and recursive self-improvement, solving long-horizon credit assignment and enabling continual learning
Highlights
LLM Reasoning & Agent — Training Stability and Efficiency
- RL Collapse — demystifying collapse from the training-inference mismatch
- Trust Region Masking — stable long-horizon LLM reinforcement learning
- LR Schedule Fix — training-inference mismatch as an optimization problem
- Dynamic Vocabulary Pruning — taming the tail for stable LLM-RL
- Optimal Token Baseline — variance reduction for long-horizon LLM-RL
- SimpleTIR — end-to-end RL for multi-turn tool-integrated reasoning
- Dr. Kernel — reinforcement learning done right for Triton kernel generation
Scalable Exploration
- HyperAgent — scalable posterior sampling bridging theory and practice
- Ensemble++ — scalable exploration via neural ensemble methods
Earlier work on deep RL and RL theory — see full CV for details.
Contact
szrlee [at] gmail [dot] com