Hi, I'm Shuyao. Chasing AGI with efficiency and agency.

Intern at Kimi; Incoming MSCS at Stanford. I explore two frontiers: improving the efficiency of singular models and designing agentic systems that are efficient and general.

Email · GitHub · X (Twitter) · Huggingface

Shuyao Xu

Experience

Kimi

Mar 2026 - Present
Intern · RL Team
  • Working on agents.

TikTok

Jun 2025 – Jan 2026
Machine Learning Engineer Intern
  • Improved account search relevance.

INF AI

Dec 2024 – May 2025
Research Intern · Host: Dr. Weidi Xu
  • Post-trained INFLogic-32B-RL via online RL. SOTA on ZebraLogicBench (85.1%).

Research

Area Venue Title
Efficiency ACL 2026 Main Reinforcement Distillation from Teacher Data for LLM Reasoning Shuyao Xu, C. Peng, J. Long, W. Xu, W. Chu, Y. Qi Efficiency ICLR 2025 Workshop The Surprising Effectiveness of Randomness in LLM Pruning Shuyao Xu, J. Liu, Z. He, C. Peng, W. Xu Efficiency Research Blog Stratiformer: Depth-Aware Heterogeneous Layer Design for Efficient Language Models Agency Research Blog Fully Agentic Test-Time Scaling Agency Research Blog Tournament-based Test-Time Scaling to Solve the Hardest Problems
Read

Education

Stanford University 2026 – 2028
M.S. in Computer Science
National University of Singapore 2022 – 2026
B.Comp in Computer Science (Honours) · GPA: 4.8 / 5.0

Teaching & Open Source

CS2103T Software Engineering TA NUS · Fall 2024 · Feedback: 4.4/5.0
MarkBind Contributor Features & mentoring junior contributors