Ali Asad | Queen's University

About Me

I am a Ph.D. student in the Department of Electrical and Computer Engineering at Queen’s University, supervised by Prof. Xiaodan Zhu at the Ingenuity Labs Research Institute. I am also a Vector Institute Scholar.

My research focuses on a defining challenge of our time: How can increasingly capable AI systems remain safe, controllable, and aligned with human values? I combine empirical and theoretical approaches to study AI safety, alignment, and controllability, examining not only where current techniques succeed but also where they may fail to scale to more capable systems.

Before starting my Ph.D., I completed my B.Sc. in Computer Engineering at Amirkabir University of Technology and spent three years in industry developing and deploying end-to-end machine learning solutions at scale for more than 15 million users.

If you are interested in my research, would like to discuss how to make increasingly capable LLMs safer, or would like to explore potential collaborations, please feel free to get in touch. I am best reached by email at ali.asad@queensu.ca.

Research Interests

AI Safety & Alignment: scalable oversight, AI deception, and mechanistic interpretability
Large Language Models & Natural Language Processing: reasoning, multi-agent systems, knowledge representation

News

[Aug 2026] I’ll be one of the speakers presenting RedDebate at the Vector Institute’s Endless Summer School.
[May 2026] Our paper RedDebate is accepted at ICML 2026.
[Dec 2025] I completed my final course at Queen’s, earning a perfect cumulative GPA.
[May 2025] Our paper IDRBench is accepted at the ICML 2025 Workshop on Assessing World Models.
[Feb 2025] I delivered a workshop on LLM reasoning and in-context learning at AAISS.
[Feb 2025] An abbreviated version of my B.Sc. thesis on knowledge graph embeddings for Persian question answering is published in Language and Linguistics.
[Sep 2024] I received the Queen’s Graduate Award for outstanding academic achievement.
[Apr 2024] I was awarded the Vector Institute Scholarship ($17,500) for excellence in AI studies.

Publications

ICML

RedDebate: Safer Responses through Multi-Agent Red Teaming Debates

Ali Asad, Stephen Obadinma, Radin Shayanfar, Xiaodan Zhu

International Conference on Machine Learning (ICML), 2026.

PDF Code Slides

Preprint

Confidently Deceptive: On the Relationship Between Confidence and Deception in LLMs

Ali Asad, Stephen Obadinma, Anshul Pattoo, Wenxuan Zhang, Xiaodan Zhu

arXiv preprint arXiv:2607.20444

PDF Code

ICML-W

IDRBench: Understanding the Capability of Large Language Models on Interdisciplinary Research

Yuanhao Shen, Daniel de Sousa, Ricardo Nascimento, Ali Asad, Hongyu Guo, Xiaodan Zhu

ICML 2025 Workshop on Assessing World Models, 2025.

PDF Code

Services

Conference Reviewers

International Conference on Learning Representations (ICLR) 2027