Sasha Cui

Ph.D. candidate, Department of Statistics & Data Science, Yale University

A.B. Mathematics, 2023, University of Chicago

I'm interested in Large Language Models and Artificial Intelligence.

Keywords: Unsupervised Activation Steering, (Multi-)Agent Benchmarking, Mechanistic Intepretability

Mail: sasha dot cui at yale dot edu

Kline Tower, Room 1037
219 Prospect St
New Haven, CT 06511

Working Group

In the Becoming Full Stack AI Researchers working group, we cover commonly used packages for serving, evaluating, and steering LMs & Agents.

Works

PAS (Painless Activation Steering) [arxiv] [github]
ABC (Agentic Benchmark Checklist) [arxiv] [lesswrong] [github] [front page of hackernews] first place at Berkeley AgentX Summit (Benchmarks & Evaluations) NIPS2025
T-Bench (Unifying Entrypoint for Coding Eval) [leaderboard] [github]

Call for Collaborators

I'm looking for collaborators who appreciate The Bitter Lesson, have good tastes, and aim to tackle AI research problems with the potential to shape the world. I work with people from all backgrounds, including undergraduate students, graduate students, & professionals. Please reach out if you're interested.

Teaching

A list of courses for which I served as the Teaching Assistant follows.

SDS540: Probability, Spring 2025
SDS542: Theory of Statistics, Fall 2024
MATH112: Calculus of Functions of One Variable, Fall 2023
Directed Reading on Machine Learning, Fall 2023
MATH240: Introduction to Applied Math, Spring 2021
MATH501: Abstract Algebra, Fall 2020