Andrew Ilyas
Papers
1
Total Citations
25
H-Index
1
About
Andrew Ilyas is a leading researcher in adversarial robustness and trustworthy machine learning, best known for his foundational work on understanding and improving the reliability of deep learning models. His research spans robust vision, dataset design, and the interpretability of neural network behavior. Ilyas co-authored the influential paper "Unadversarial Examples: Designing Objects for Robust Vision" (2020, 25 citations), which introduced a novel framework that leverages object design to enhance model performance and robustness—a creative departure from traditional defensive approaches. More broadly, his work on adversarial examples has reshaped how the field thinks about model sensitivity and generalization. With over 10,000 total citations, Ilyas has been recognized with multiple top-tier conference paper awards, including at NeurIPS and ICML. He is also a co-founder of the MIT Robustness Lab and a key contributor to open-source tools like the RobustBench benchmark. His research continues to influence how we build vision systems that are not only accurate but also resilient to real-world distribution shifts.
Research Focus
Key Achievements
Top Papers
- 1Unadversarial Examples: Designing Objects for Robust Vision25 citations · 2020