Best Graduate Research Assistant
Wayne State University, 2025
Ph.D. Candidate · Computer Science
Saleh Zare Zade
I study how large language models remember, forget, and fail. I build methods for safer, more private, and more trustworthy AI.
About
I am a Ph.D. candidate in Computer Science at Wayne State University, advised by Prof. Dongxiao Zhu.
My research focuses on large language model safety, privacy, and unlearning, with an emphasis on membership inference attacks, memorization behavior, and robust mitigation strategies for trustworthy AI systems.
Selected work
Recent publications
Research spanning machine unlearning, privacy attacks, and the security of language models.
-
01
Attention Smoothing Is All You Need For Unlearning
Zare Zade, S., Zhou, X., Liu, S., Zhu, D.
-
02
Automatic Calibration for Membership Inference Attack on Large Language Models
Zare Zade, S., Qiang, Y., Zhou, X., et al.
-
03
Not All Tokens Are Meant to Be Forgotten
Zhou, X., Qiang, Y., Zare Zade, S., et al.
Recognition
Awards & service
Academic recognition and contribution to the research community.
Undergraduate distinction
Ranked #4 with direct entry to the Master’s program.
Academic service
Reviewer for ICML 2026 (Gold Reviewer), NeurIPS 2026, ICMLA 2026, AAAI 2027, and ICLR 2027.
Contact