Kerem Zaman

PhD student at UNC Chapel Hill

prof_pic.jpg

UNC Chapel Hill

NC, USA

I’m currently a PhD candidate at the UNC-Chapel Hill where I am advised by Shashank Srivastava. My research lies at the intersection of interpretability and AI safety. I combine black-box and white-box approaches to develop methods for monitoring LLM systems. My broader goal is to build safer and more transparent AI systems by deepening our understanding of how black-box models behave and what drives their behavior.

In summer 2026, I was a Student Researcher at Google DeepMind NYC, working with Neha Kalibhat and Mani Malek. Previously I interned at IFM MBZUAI Silicon Valley Lab with Mikhail Yurochkin. Before PhD, I earned my bachelor’s degree from the Boğaziçi University, where my thesis advisor was Pinar Yanardag.

news

  • Paper accepted into ACL 2026 as oral! “Is Chain-of-Thought really “not explainability”? [pdf] [code]
  • This summer I’ll be interning at Google DeepMind NYC Office!
  • Paper accepted into ICLR 2026! “DiffVax: Optimization-Free Image Immunization Against Diffusion-Based Editing” [pdf] [website]
  • New preprint! Is Chain-of-Thought really “not explainability”? [pdf]
  • Selected as an Outstanding Reviewer at EMNLP 2025 (<1%)!

selected publications