Zedian (Mark) Xiao
ML Researcher & Engineer

I am learning to dream big and to turn those dreams into reality. My current research interests are AI for knowledge discovery and knowledge injection in LLMs. Previously, I showed how to scale LLMs to learn from millions of human workflows and how to hot-swap knowledge with LoRA adapters.

I have obtained my MS in Computer Science from Stanford University, advised by Diyi Yang. And my BS in Mathematics and Computer Science from McGill University, advised by Xue (Steve) Liu and Masoud Asgharian.

In my free time, you can find me tinkering with my espresso machine and discovering that red wines are not all the same over dinner with friends.

Happy to chat! 😊 → markxiao[at]stanford.edu

Dinner with friends: beef Wellington and glasses of red wine

Selected publications

ScribeAgent: Towards Specialized Web Agents Using Production-Scale Workflow Data Zedian Xiao, Atishay Jain, Junhong Shen, Ishan Amlekar, Mouad Hadji, Aaron Podolny, Ameet Talwalkar ICLR 2025 · code
HyperLoRA: Task-Agnostic Low-Rank Adapters for Unseen English Dialects Zedian Xiao, William Held, Yanchen Liu, Diyi Yang EMNLP 2023 · arXiv · code

Mentors and Friends I learned a lot from

Diyi Yang / Jie Fu / Xue (Steve) Liu / Masoud Asgharian / Vahid Partovi Nia / Atishay Jain

Teaching

McGill University

COMP250 Data Structures
COMP251 Algorithms
COMP302 Programming Languages
COMP360 Advanced Algorithms
MATH203 Principles of Statistics I
MATH204 Principles of Statistics II
MATH235 Abstract Algebra
MATH236 Linear Algebra
MATH242 Real Analysis
MATH323 Probability

Stanford University

CS229 Machine Learning
CS236 Deep Generative Models
CS261 Combinatorial Optimization

Service

ACM CAIS 2026 Reviewer and artifact reviewer