Hi, I’m Beyza.
I work at the intersection of AI safety and alignment, with a particular interest in the reliability of multi-agent interactions: how AI systems behave, influence one another, and stay trustworthy as they become more autonomous.
Contact: nbozdag2 [at] illinois [dot] edu
News
Publications
Highlighted Work
Too Polite to Disagree: Understanding Sycophancy Propagation in Multi-Agent Systems
Few-Shot Synthetic Accented Speech for ASR Fine-Tuning: What Helps and When?
Research Experience
Applied Scientist Intern
Applied Scientist Intern
Graduate Research Assistant
Education
Ph.D. in Computer Science
Conversational AI Lab, advised by Dilek Hakkani-Tür and Gokhan Tur.
B.S. in Computer Science & Mathematics
Undergraduate researcher, CLU Lab, advised by Steve Bethard.
Talks & Presentations
Introducing PMIYC: A Framework for Evaluating Persuasion Effectiveness and Susceptibility
Blog
To Be or Not to Be Persuaded: Persuasion in Large Language Models
How well can large language models persuade, and how easily are they persuaded in turn? A look at both sides of persuasion in LLMs, and why it matters for AI safety.
Read more →Outside the Lab
I like to wind down by painting sometimes, mostly with oil paint, and lately I've been getting into watercolor too!
Traveling, sightseeing, hiking: the world has too many good views to spend all day in front of a computer. At least now my coding agent can handle the busy work :)
A little goofy never hurt anyone. This is a parody-song channel (Two Brains, One Breakdown) I run with my dear friend and labmate, Ishika. We rewrite popular songs into odes to PhD life (deadlines, reviewer #2, and everything in between), and make other fun content too!