5th Year Master of Science Thesis Defense - Michael Cui
— 12:30pm
Location:
In Person
-
Reddy Conference Room, Gates Hillman 4405
Speaker:
MICHAEL CUI
,
Master's Student, Computer Science Department, Carnegie Mellon University
Assigning submitted papers to appropriate reviewers is a fundamental component of the peer-review process in large academic conferences. In modern conference settings, this task has become increasingly challenging due to the scale of submissions and the need to satisfy multiple competing objectives simultaneously. In particular, program chairs must balance reviewer expertise, diversity considerations, and robustness to strategic behavior, while also ensuring that the assignment process remains practical at scale. This thesis aims to improve the paper-assignment process by making it more effective, more robust, and more practical for real-world peer review.
This thesis studies the problem of large-scale paper assignment from an optimization perspective. It examines the limitations of existing assignment methods, which often optimize only a subset of the relevant objectives or become computationally impractical in realistic conference settings. To address these limitations, the thesis presents Robust Assignment via Marginal Perturbation (RAMP), a unified framework for scalable, robust, and diversity-aware reviewer assignment. The proposed framework combines a linearized perturbed-maximization objective with soft constraints that incorporate multiple practical desiderata into a single optimization procedure, together with an attribute-aware sampling method for converting fractional assignments into integral ones.
In addition to presenting the algorithmic framework, this thesis discusses the practical challenges and lessons that arose in deploying the method for major AI conferences, including AAAI 2026, AAMAS 2026, and EC 2026. It also describes an interface that enables future conference organizers to run the matching process directly, helping bridge the gap between optimization research and real conference workflows.
Thesis Committee
Fei Fang (Chair)
Nihar Shah
Additional Information
For More Information:
amalloy@cs.cmu.edu