About
Hi there! I am a founding engineer of RadixArk, a CS PhD student, a proud member of lmsys.org, and a modest learner of Machine Learning Systems. I live near Los Angeles with my girlfriend and two cats.
I feel honored to be part of SGLang team and maintain SGLang-Omni for multi-stage model inference optimization. I always welcome self-motivated engineers and researchers to join us, and try our best to refer new opportunities.
I spent one year at UCLA as a PhD student, fortunately advised by Prof. Quanquan Gu, Dr. Ying Sheng and Dr. Lianmin Zheng.
I obtained my Bachelor's degree at Tsinghua University. At that time, I was supervised by Prof. Graham Neubig and Prof. Tongshuang Wu and mentored by Vijay Viswanathan at CMU. Their research guidance and tastes are invaluable to me.
Research
Post-training Pipelines
Multi-stage Model Inference
Proud Publications
SwingArena: Competitive Programming Arena for Long-context GitHub Issue Solving
RLVE: Scaling Up Reinforcement Learning for Language Models with Adaptive Verifiable Environments
Soft Thinking: Unlocking the Reasoning Potential of LLMs in Continuous Concept Space
MiniCPM: Unveiling the Potential of Small Language Models with Scalable Training Strategies
Internet of Agents: Weaving a Web of Heterogeneous Agents for Collaborative Intelligence
SELF-GUIDE: Better Task-Specific Instruction Following via Self-Synthetic Finetuning
Prompt2Model: Generating Deployable Models from Natural Language Instructions
Resume
Hobbies
Steel-string Acoustic Guitar 民谣吉他
Swimming, Running and Hiking
Cooking Szechwan Cuisine 川菜